การประชุมแถลงผลประกอบการไตรมาส 2 ปี 2026 ของ Globant (GLOB): ARR ของ Glob.AI พุ่งสูงขึ้น, ปรับลดประมาณการรายได้

ประเด็นสำคัญ

• รายได้ในไตรมาส 2 ปี 2026 อยู่ที่ 614.4 ล้านดอลลาร์ เพิ่มขึ้น 1.2% เมื่อเทียบกับไตรมาสก่อนหน้า และสูงขึ้นเล็กน้อยเมื่อเทียบกับช่วงเดียวกันของปีก่อน โดยได้รับปัจจัยหนุนจากผลกระทบเชิงบวกของอัตราแลกเปลี่ยน 80 basis points ซึ่งช่วยสนับสนุนการเปรียบเทียบแบบรายปี
• รายได้ประจำเป็นรายปี (ARR) ของ Glob.AI แตะระดับ 52.8 ล้านดอลลาร์ ณ เดือนมิถุนายน เพิ่มขึ้นประมาณ 60% จาก 32.8 ล้านดอลลาร์ในเดือนมีนาคม โดยผู้บริหารคาดว่ายอดดังกล่าวจะทะลุ 110 ล้านดอลลาร์ภายในสิ้นปี 2026
• AI Pods ได้รับการนำไปใช้โดยลูกค้ารายใหญ่ 45 ราย และคิดเป็น 45% ของลูกค้ารายใหญ่ที่สุด 20 อันดับแรกของ Globant โดยมีอัตรากำไรขั้นต้นสูงกว่าอัตรากำไรของการส่งมอบรูปแบบเดิมเกือบ 10 percentage points
• รายได้ต่อพนักงานปรับตัวขึ้นสู่ระดับ 95,800 ดอลลาร์ต่อปี เมื่อคิดคำนวณแบบปรับเป็นอัตราต่อปี (Annualized run rate) ซึ่งเพิ่มขึ้น 9.7% เมื่อเทียบกับช่วงเดียวกันของปีก่อน ขณะที่จำนวนพนักงานทั้งหมดลดลงเกือบ 10%
• Globant ปรับลดคาดการณ์รายได้ปี 2026 ลงมาอยู่ที่ 2.428 พันล้านดอลลาร์ - 2.462 พันล้านดอลลาร์ โดยอ้างถึงความล่าช้าของโครงการในตลาดใหม่ แรงกดดันต่อลูกค้าในกลุ่มการท่องเที่ยว และวัฏจักรการตัดสินใจใช้จ่ายที่ไม่จำเป็นที่ยาวนานขึ้น
• โครงการปรับโครงสร้างธุรกิจเพื่อเพิ่มประสิทธิภาพส่งผลให้เกิดค่าใช้จ่ายครั้งเดียวจำนวน 32.3 ล้านดอลลาร์ในไตรมาส 2 โดยผู้บริหารคาดว่าจะมีค่าใช้จ่ายที่เกี่ยวข้องอีก 20 ล้านดอลลาร์ - 25 ล้านดอลลาร์ในไตรมาส 3

ข้อมูลทางการเงินที่สำคัญ

ผลการดำเนินงานทางธุรกิจและการปฏิบัติงาน

Globant กำลังปรับเปลี่ยนรูปแบบการส่งมอบงานส่วนหนึ่งจากการคิดค่าบริการตามเวลาและใช้วัสดุ (Time-and-materials) หรือรูปแบบขอบเขตงานคงที่ ไปสู่ AI Pods ที่กำหนดราคาตามผลงาน คุณค่า หรือการบริโภคใช้จริง โดยรายได้จาก AI Pods คิดเป็นประมาณ 2% ของรายได้ทั้งหมดในไตรมาส 2 ซึ่งผู้บริหารคาดว่าสัดส่วนดังกล่าวจะเข้าใกล้ 4% เมื่อคิดคำนวณตามอัตราการเติบโต ณ สิ้นปี

บริษัทรายงานว่า AI Pods มีผลิตภาพสูงกว่าการใช้วิธีวิศวกรบวกกับ AI แบบทั่วไปมากกว่า 30% โดยผู้บริหารกล่าวว่าอัตรากำไรขั้นต้นของรูปแบบนี้สูงกว่าการส่งมอบแบบเดิมเกือบ 10 percentage points ซึ่งสนับสนุนการตัดสินใจเร่งเปลี่ยนผ่านลูกค้า แม้ว่าจะอาจเกิดแรงกดดันต่อรายได้ในระยะสั้นก็ตาม

Glob.AI ซึ่งเปิดตัวก่อนการแถลงผลประกอบการไม่นาน ให้บริการเข้าถึง AI Pods ผ่านแพลตฟอร์มแบบบริการตนเอง (Self-service) สถาปัตยกรรมของแพลตฟอร์มสามารถส่งต่องานไปยังโมเดลภาษาขนาดใหญ่ (LLM) มากกว่า 140 โมเดล พร้อมอนุญาตให้ลูกค้าควบคุมการเลือกและการใช้งานโมเดลได้เอง โดย Globant ได้เน้นย้ำถึงความร่วมมือกับ Anthropic, OpenAI, Vercel, NVIDIA, Salesforce และ MuleSoft

Data and AI Studio ได้กลายเป็นสตูดิโอที่ใหญ่ที่สุดเป็นอันดับสองของ Globant โดยสร้างยอดขายเกือบ 11% และเติบโตประมาณ 35% เมื่อเทียบกับช่วงเดียวกันของปีก่อน ทั้งนี้ ลูกค้าเดิมคิดเป็น 96% ของรายได้ประจำไตรมาส

กลุ่มลูกค้ารายใหญ่มีผลการดำเนินงานโดดเด่นกว่าค่าเฉลี่ยของบริษัท โดยรายได้จากลูกค้ารายใหญ่ 50 อันดับแรกเพิ่มขึ้น 6.9% เมื่อเทียบกับช่วงเดียวกันของปีก่อน ขณะที่กลุ่ม 20 อันดับแรกและ 10 อันดับแรกเติบโต 6.6% และ 4.4% ตามลำดับ นอกจากนี้ ลูกค้า 16 รายจาก 20 อันดับแรกมีการขยายความร่วมมือเพิ่มขึ้นเมื่อเทียบกับช่วงเดียวกันของปีก่อน

ผลการดำเนินงานในแต่ละภูมิภาคมีความหลากหลาย โดยยุโรปเติบโต 6.8% และละตินอเมริกาเพิ่มขึ้น 5.9% เมื่อเทียบกับช่วงเดียวกันของปีก่อน ขณะที่อเมริกาเหนือหดตัว 2.4% และตลาดใหม่ลดลง 17.7% เนื่องจากความล่าช้าของโครงการที่ไม่คาดคิด ทั้งนี้ ตลาดใหม่ได้ฉุดการเติบโตรวมแบบรายปีลงประมาณ 115 basis points

คาดการณ์โดยผู้บริหาร

ช่วงการคาดการณ์รายได้ตลอดปีเดิมอยู่ที่ 2.462 พันล้านดอลลาร์ - 2.508 พันล้านดอลลาร์ โดยผู้บริหารกล่าวว่าค่ากลางที่ปรับใหม่ซึ่งอยู่ที่ประมาณ 2.445 พันล้านดอลลาร์ ส่งผลให้รายงานการเติบโตลดลง 0.4% และลดลง 1.1% หากคำนวณด้วยอัตราแลกเปลี่ยนคงที่ ซึ่งรวมถึงปัจจัยหนุนจากอัตราแลกเปลี่ยนที่คาดว่าอยู่ที่ประมาณ 70 basis points

ผู้บริหารระบุว่าประมาณครึ่งหนึ่งของการปรับลดคาดการณ์มาจากความคาดหวังที่ลดลงในตลาดใหม่ ขณะที่อีกประมาณ 10 ล้านดอลลาร์เชื่อมโยงกับแรงกดดันต่อลูกค้าในกลุ่มการท่องเที่ยวและการบริการ ส่วนที่เหลือสะท้อนถึงข้อสมมติฐานการเปลี่ยนผ่านและความต้องการในภาพรวมของตลาด

Globant คาดว่าจะมีกระแสเงินสดอิสระที่แข็งแกร่งในครึ่งปีหลัง ซึ่งสอดคล้องกับปัจจัยฤดูกาลตามปกติ โดยลำดับความสำคัญในการจัดสรรเงินทุนยังคงเป็นโครงการซื้อหุ้นคืนมูลค่าสูงสุด 125 ล้านดอลลาร์ในระยะเวลาหกไตรมาส และการลงทุนอย่างต่อเนื่องใน AI Pods และ Glob.AI

ความเสี่ยงและประเด็นที่ต้องจับตา

• แรงกดดันทางภูมิรัฐศาสตร์และโครงการที่ล่าช้าส่งผลให้รายได้ในตลาดใหม่ลดลง 17.7% ในช่วงไตรมาส 2
• ความผันผวนของราคาน้ำมันส่งผลให้ลูกค้าในกลุ่มการท่องเที่ยวบางรายชะลอโครงการปรับเปลี่ยนองค์กรเพื่อปกป้องความสามารถในการทำกำไร
• ลูกค้าในอเมริกาเหนือยังคงแสดงให้เห็นถึงวัฏจักรการตัดสินใจที่ยาวนานขึ้นสำหรับการใช้จ่ายที่ไม่จำเป็น
• การอ่อนค่าของดอลลาร์สหรัฐส่งผลให้ต้นทุนในศูนย์ส่งมอบงานในละตินอเมริกาสูงขึ้น โดยผู้บริหารประเมินว่าความเคลื่อนไหวของสกุลเงินในภูมิภาคสร้างผลกระทบต่ออัตรากำไรโดยรวมประมาณ 4 percentage points ในช่วง 18 ถึง 24 เดือนที่ผ่านมา
• การเร่งการเปลี่ยนผ่านไปสู่ AI Pods อาจลดรายได้จากสัญญาบางฉบับ หาก Globant แบ่งปันผลประหยัดที่เกิดจากผลิตภาพในทันทีให้แก่ลูกค้า
• AI Pods ยังคงเป็นเพียงสัดส่วนเล็กน้อยของรายได้ทั้งหมด แม้ว่า ARR จะเติบโตอย่างรวดเร็วและมีอัตรากำไรที่รายงานสูงขึ้นก็ตาม

ประเด็นสำคัญจากการถาม-ตอบกับนักวิเคราะห์

ผู้บริหารกล่าวว่าการเปลี่ยนผ่านไปสู่ AI-native จะดำเนินไปอย่างรอบคอบในแต่ละไตรมาส และอาจส่งผลกระทบอย่างมีนัยสำคัญต่อการเติบโตของบริษัทมากขึ้นภายในสิ้นปีหน้า พร้อมเน้นย้ำว่าเหตุผลหลักในการปรับคาดการณ์ปี 2026 คือความอ่อนแอในตลาดใหม่ มากกว่าการเปลี่ยนผ่านไปสู่ AI Pod

Globant ระบุว่าขั้นตอนการอนุมัติตามกระบวนการจัดซื้อจัดจ้างอาจชะลอการเปลี่ยนไปสู่สัญญาที่อิงตามผลงานหรือการบริโภคใช้จริง ในบางกรณี บริษัทอาจยอมรับรายได้เริ่มต้นที่ต่ำลงเพื่อเร่งการยอมรับและใช้งาน ขณะเดียวกันก็พยายามรักษาความสามารถในการทำกำไรผ่านเศรษฐศาสตร์ต่อหน่วย (Unit economics) ที่ดีขึ้น ขยายไปสู่แผนกอื่น ๆ ของลูกค้า และแย่งชิงงานจากคู่แข่ง

ผู้บริหารอธิบายถึงจุดเด่นที่แตกต่างของ Glob.AI ว่าเป็นการผสมผสานระหว่างความเป็นอิสระของโมเดล กระบวนการทำงานที่ทำซ้ำได้ การกำกับดูแลโดยมนุษย์ และความรับผิดชอบต่อผลลัพธ์ทางธุรกิจ ทั้งนี้ แพลตฟอร์มได้รับการออกแบบมาเพื่อลดการสูญเสียโทเคน (Token waste) และผลลัพธ์ที่ไม่สม่ำเสมอ โดยการกำหนดอินพุต เอาต์พุต และการควบคุมคุณภาพที่จำเป็นสำหรับแต่ละกระบวนการทำงาน

ผู้บริหารยังกล่าวอีกว่า Glob.AI สามารถขยาย Globant ให้เติบโตเกินกว่าการพัฒนาซอฟต์แวร์ ไปสู่การปฏิบัติงานทางธุรกิจที่คิดราคาตามธุรกรรมหรือผลลัพธ์ โดยบริการทางการเงิน สื่อและความบันเทิง และสายการบิน เป็นกลุ่มอุตสาหกรรมที่เริ่มเห็นสัญญาณตอบรับในเบื้องต้น

โครงการเพิ่มประสิทธิภาพรวมถึงการยกระดับทักษะแรงงาน (Reskilling) การควบรวมสำนักงาน และการจัดลำดับความสำคัญของศูนย์ส่งมอบงาน โดย Globant ตั้งใจที่จะใช้กำลังความสามารถด้านต้นทุนที่ได้รับจากการนี้บางส่วนเพื่อชดเชยแรงกดดันด้านสกุลเงิน และสนับสนุนทุนในการพัฒนา Glob.AI ตลอดจนบุคลากรด้าน AI

บันทึกข้อความการแถลงผลประกอบการฉบับเต็ม

บทถอดเสียงฉบับเต็มของการประชุมทางโทรศัพท์ผลประกอบการ

คำชี้แจงจากฝ่ายบริหาร

Arturo Langa

[Presentation]

Arturo Langa

Good afternoon, and welcome to Globant's Second Quarter 2026 Earnings Conference Call. I am Arturo Langa, Investor Relations Officer at Globant. [Operator Instructions] Please note, this event is being recorded and streamed live on YouTube.

By now, you should have received a copy of the earnings release. If you have not, a copy is available on our website, investors.globant.com.

We will begin with remarks by our Chief Executive Officer, Martin Migoya; our Chief Technology Officer, Diego Tartara; and our Chief Financial Officer, Juan Urthiague; followed by a Q&A, where they will be joined by our Chief Revenue Officer, Fernando Matzkin.

Before we begin, I would like to remind you that some of the comments on our call today may be deemed forward-looking statements. This includes our business and financial outlook and the answers to some of your questions. Such statements are subject to the risks and uncertainties as described in the company's earnings release and other filings with the SEC.

Please note that we follow IFRS accounting rules in our financial statements. During our call today, we will report non-IFRS or adjusted measures, which is how we track performance internally and the easiest way to compare Globant to our peers in the industry. We'll find a reconciliation of IFRS and non-IFRS measures at the end of the press release we published on our Investor Relations website announcing this quarter's results.

I will now turn the call over to Martin Migoya.

Martín Migoya

Good afternoon, everyone, and thank you for joining us. Today, I want to talk about a change we are leading, a new way of creating value for our customers, delivering our work, a way of pricing it, which is already starting to compound. For more than 20 years, we have engineered the digital reinvention of the world's leading organizations, building the software products and platforms that run their businesses, delivered by dedicated high-performing teams and priced through on fixed scope engagements or time and materials. That work remains the backbone of Globant.

One year ago, I introduced you to AI Pods, a new AI native revenue stream built on that foundation, but priced on the actual output and value we deliver or on consumption, rather than on the hours we bill. As AI Pods deliver more work at higher margin for a similar price, our top line can understate the progress underneath it. As this grows, it will be relevant to assess annual recurring revenue, revenue per head, AI Pod margins and client penetration, alongside the total revenue line.

Before I go further, let me be precise about 2 names you will hear all call. Glob.AI is the platform we opened to the market last week. AI Pods are the service units that live on it, run by AI agent workflows and supervised by our experts. The revenue they create, I will call Glob.AI ARR. Hold those 3 together: the platform, the pods and the number.

This quarter's revenues grew roughly 60% to $52.8 million, and we estimate that Glob.AI's ARR will surpass $110 million by year-end. Let me walk you through it in that order. The model, the number that measures it, where the growth is coming from, how to read our reported revenue, while both models run side by side and where it already shows results. The technology services industry as a whole is growing at roughly flat rates right now but flat is an average, and averages hide the real story. Inside that flat industry, we have found a growth runway: AI native services. And it is growing because it expresses what enterprises want for AI better than a traditional hours-based approach does. Clients can tell the difference, and a growing number are moving budgets accordingly and more are choosing to work with us with this new model.

As enterprises abstract away layer after layer of complexity, infrastructure, platform, software, some are now beginning to abstract away business services themselves. We think of this as service as software. Just like how the cloud transformed software infrastructure and provided predictable and recurring revenue, Glob.AI does for professional services. You turn on the outcome and pay for what you consume with Globant's experts built in.

It opens budgets we have not had access to before. Annual spending of the global professional services industry is estimated at more than $6 trillion, roughly 4x the size of the IT services market that Globant has been evaluated in. We are taking this deliberate decision to respect our current market while expanding our offering to a larger total addressable market. We are steering clients toward the new model. Our AI-native delivery system of AI Pods has now been adopted by 45 of our clients in many of their projects.

Today, practically everything we deliver carries AI. We do not count that as Glob.AI ARR, which only captures the revenue that is delivered and charged differently on the output, value or consumption our clients receive, not on the hours behind it. It is a strict measure and that is deliberate. When this number grows, it is not AI being bolted on to existing work. It is the business model itself changing.

Glob.AI ARR reached $52.8 million as of June, up from $32.8 million in March, roughly 60% growth in a single quarter. We have seen plus 30% more productivity than with a typical engineer plus AI approach. Pipeline stands at $436.8 million, up from $352 million in Q1. Adoption has reached 45% of our top 20 accounts. Gross margins on this model run close to 10 percentage points above our traditional delivery. We now expect to exit 2026 at no less than $110 million in Glob.AI ARR. That is the yardstick, AI-native revenue becoming core to how we believe the market should value Globant, measured quarter after quarter.

These changes have affected Globant as a whole as well. Globant's revenue per head reached $95,800 on a run rate basis, up 9.7% year-over-year. We are delivering more value with the same talent and capturing it. Glob.AI ARR captures 3 of the biggest waves of demand in our industry. They are core modernization, experienced debt and agentic process transformation. We have shared them with you on previous earnings calls. What changed is that all 3 now convert increasingly through AI Pods.

Let me go through each one with you. One, core modernization. There is a technical debt backlog between $1.5 trillion and $2 trillion across the world's 2,000 largest public companies. It used to mean a large team billing hours over months. We can now deliver it as an outcome in less time. That is why clients are moving to the new model here first.

Two, experienced debt. Every customer-facing surface that has to be rebuilt for an AI-first world, our Vercel and cloud-powered AI Pods are turning multi-month rebuilds into same week releases. Three, agentic process transformation, the largest opportunity, redesigning how our business runs around agents. The value is in the transformed process, not the hours. So this is where outcome pricing fits best.

Now a word on our current position and how to interpret the top line while this shift is underway. For Q2, revenue was $614.4 million, within our guided range, up 1.2% sequentially and back to slight year-over-year growth. AI Pod revenue makes up roughly 2% of our total revenue today, and we expect it to reach 4% by the end of the year.

We keep seeing the pocket of growth I mentioned earlier, demand for AI Pods. We are choosing to accelerate these migrations, even if it means a short-term impact on revenue. Because over time, it creates more value for the client with predictable outcome-based consumption and more value for Globant with higher margins and access to more sophisticated projects.

This quarter, 96% of our revenue came from repeat customers, and we grew our top 20 and top 50 clients by 6.6% and 6.9% year-over-year, respectively. This is concentrated where Glob.AI and AI Pod penetration is highest. Our Data and AI Studio is now our second largest studio by revenue, close to 11% of sales and growing close to 35% year-over-year. Our AI studios are increasingly selling AI-native services alongside traditional staff augmentation, and providing their depth, since that top 50 growth is a sign they understand these clients' industry as well.

Our core business is acting as the distribution engine that carries Glob.AI, and the AI pods that run on it into large enterprises on relationships built over 2 decades. Pipeline and bookings are at a healthy level. Its composition is shifting towards AI, data, cloud and integration work.

Having said all this, we are operating in a tougher environment this quarter: Geopolitical pressure in our new markets, volatile oil prices weighing on travel, and longer decision cycles in North America. Juan will take you through a revised outlook for the full year.

Last week, we launched Glob.AI. And with that, we are opening the same model to any enterprise through a single self-service platform, so AI-native services, priced on output and value or on consumption, become available to more of the market, not just our largest accounts.

Here is what that looks like in practice. Glob.AI is a single destination where an enterprise can find, deploy and start consuming an AI Pod without a months-long discovery process and a long ramp-up time. A client can log in, explain their technological opportunity and plan language, and the platform draws on Globant's entire network of technological solutions, partnerships, recommends AI Pods and enables clients to start building the same day. It bridges the gap between mental throughput and making sound business decisions.

Clients keep sovereignty over which models they use and where they run. These pods are built in cooperation with the companies defining this technology. Specific AI Pods are engineered with name partners, secure code review with Anthropic, digital twin engineering on NVIDIA Omniverse, prototype-to-product with Vercel and enterprise integration with Salesforce and MuleSoft. The platform runs across the broader model ecosystem as well: Anthropic, OpenAI, Google, Azure, AWS, NVIDIA, Meta, among others.

I am glad to announce that Sarab Narang is joining us as Glob.AI's CEO. Sarab is an accomplished AI and technology executive with more than 23 years of experience. He joins us from ServiceNow, where he led the commercialization of its AI business and previously held senior AI leadership roles at AWS where he helped build and scale AI platforms, including Amazon SageMaker and Amazon Bedrock. Earlier in his career, he built KPMG's AI and machine learning practice.

None of what I have discussed so far works without the right partners. In June, we announced a multiyear alliance with Anthropic, becoming a preferred services partner in the Claude Partner Network. Since signing, we have moved quickly. Several Claude-powered AI pods are already in production and were showcased at our Globant Tech Summit in July. We are training thousands of Globers and Anthropic tools, and we have an active joint pipeline with several large financial institutions, airlines and e-commerce companies. One year after our initial partnership OpenAI has named Globant a selected partner in its new partner network.

And with Vercel, clients can ship AI built applications natively in a single click, turning multi-month projects into same week deliveries. Together, we launched Vercel-powered AI Pods, agentic units that design, develop and modernize enterprise digital products on [ Next.js ].

FIFA is using AI Pods powered by Glob.AI to scale its digital ecosystem into a continuous, personalized experience for football fans worldwide. Its key platforms recognize fan preferences across competitions, and powered by AI Pods, use real-time data to generate new experiences year-round.

This quarter marked 3 years since the foundation of our partnership with British Airways, delivering a platform built for speed and continuous innovation. In June, British Airways reached an important milestone on this transformation journey with the launch of their new mobile app, following extensive testing to make every stage of the customer journey simpler and more intuitive, acting as a real-time travel companion. Positive customer feedback has highlighted the improved user experience, particularly the live flight notifications feature.

And this is just the start. We're extending the partnership with new features powered by our AI Pods model, accelerating what we can deliver next. In the Gulf region, we are working with one of its largest financial institutions by building its first agentic bank. Powered by our AI Pods model, intelligent agents will act across acquisition, onboarding, servicing and risk, reshaping how the bank operates and how customers experience it.

GUT delivered a solid Q2 2026, culminating in another standout performance at the Cannes Lions International Festival of Creativity in June. The network earned 22 Lions, including a third consecutive Grand Prix for long-standing client, Mercado Libre, the first agency client partnership to achieve this milestone at the festival. GUT also launched new work for Google Chrome and Ray-Ban Meta, created the world's first clay bar for Stella Artois at Roland Garros during the French Open, and introduced Rimowa's For a Lifetime of Lives Platform, celebrating craftsmanship through stories of longevity and evolution.

We are building a meaningfully different services business deliberately with a growing base of revenue underneath it. And with a number I have asked you to hold us to every quarter. This next chapter also means disciplined choices today, including decisions on our cost base to fund the transition and protect our margins. I do not take those lightly, and I'm grateful to our teams for the resolve they are showing.

Thank you to our clients, our partners and our Globers around the world building this alongside us every day. With that, Diego will show you the machine underneath. Thank you.

Diego Tartara

Thank you, Martin, and hello, everyone. Martin just laid out our strategic vision for Glob.AI and how it fundamentally transforms the way clients acquire our services. My focus today is on the underlying engine: the technology architecture, the operational mechanics and the first-mover advantage that makes this delivery model form at scale.

The legacy professional services model trades human hours or custom solutions, forcing engineers to solve the same foundational problems repeatedly. Glob.AI breaks that cycle by operating as an asset-based engine. Within the platform, we codified over 2 decades of enterprise engineering and industry domain knowledge into curated battle-tested playbooks. These are deterministic documented agentic workflows designed for production-grade reliability.

Because these agentic assets are modular and validated, we achieve extraordinary cross-industry compounding value. an IT root cause analysis workflow built for an airline client, for example, can be replatformed into a pharma supply chain or media distribution pipeline in a matter of weeks rather than months. AI Pods serve as our direct vehicle for monetizing this compounding IP, taking clients from a natural language challenge to production-ready deployment without the traditional friction.

As we set out in introducing Glob.AI, raw LLM prompting produces significant token waste, hallucinated logic and expensive retry loops that includes additional time for human supervision. Glob.AI solves this through structured deterministic process optimization. Before an AI agent executes work, our platform automates context assembly, architecture mapping and data preparation, enforcing automated quality gates at every step. By optimizing the orchestration layer, we ensure that every token consumed yields verifiable production-ready output.

At the same time, enterprise adoption hinges on control. Glob.AI's architecture routes intelligently across more than 140 LLMs, providing complete model independence, so clients are never locked into a single provider. Crucially, every transaction is locked within the client's dedicated token vault. This guarantees absolute token sovereignty. No client data is ever exposed or used to train third-party models, allowing organizations to compound their own institutional intelligence safely over time.

We are seeing the impact of this platform model directly in our operational performance. By integrating specialized AI agents into continuous delivery workflows supervised by our experts, we are restructuring the software engineering life cycle. This leverage enables us to decouple enhancements in output, velocity and delivery from headcount growth. Furthermore, this enables us to layer on nonlinear and recurring revenue to the mix, with structurally better unit economics that will, over time, transform the business. This is the structural signature of a business moving up the value chain.

Now I would rather show than tell. Let me take you on a short Glob.AI tour. It starts simply. Once you have selected the plan and the AI Pods from the catalog or through chat with the Glob.AI agent, you would be able to access the project's module. From that moment on, you are ready to begin.

Glob.AI works at 2 levels: your organization and your projects. Your organization is the foundation where you can figure the tools and settings that apply to all your projects. Your projects are where the work happens, each with its own goals and one or more AI pods running simultaneously at different prices, which always include not only the tokens consumed, but also the human supervision required for the project.

Let's start at the organization level. The first thing your organization needs is context, and you can give it [indiscernible] from the tools your organization already uses. Connect your GitHub to read, write and open poll request directly on your repositories; Google Workspace to pull signals from Gmail, calendar and drive; Atlassian to sync with your Jira tickets and Confluence pages. You can also provide context through a document or a description.

Once you're inside your project, everything is in one place. At the top, you see your project at a glance, resuming the totals of your project, how many AI Pods are active, your goals, your token consumption, your current spend with your limit and the start button that sets everything in motion.

Right below, your customized dashboard also gives you 4 key metrics. Selecting the period, you can see the spend, tokens used, tasks total and artifacts produced. Beneath that, a quick summary, your active goals, the units of work in progress, always visible at a glance. And finally, your AI Pod fleet, the specialized teams running your delivery.

On the right is your communication channel, the Glob.AI agent. Just describe what you need in plain language or start with one of the suggestions below. You can also mention your forward deployed engineer, the Globant expert assigned to your project and your direct line to the human overseeing every step. Whenever you need to communicate, you're one message away from your expert.

Goals are the units of work your AI Pods will execute. And you create them by simply telling the agent what you want to achieve. You can connect the repository and let the agent analyze your code base directly. You can describe your needs in plain language or you can attach a document and let the agent extract the requirements from there. Either way, the agent proposes the goals, you review them, you approve them. And if something needs to change, you tell it directly or bring in your expert, always available in the chat with full context on your project.

When you're ready, you say start. That single action locks in your project and hands it to the forward deployed engineer who supervises every step of what comes next. Artifacts is your shared workspace for project files. Here, you'll find every document, spec, [ ABR ] or implementation summary produced by your AI Pods, ready to preview with a single click. You and your FDE can all upload files here, keeping everything in one place.

Some artifacts will also be delivered directly to your GitHub repository as commits or pull requests. Clicking on your name gives you access to your notification settings. You choose how Glob.AI reaches you: e-mail, WhatsApp or Telegram. And you decide which events trigger each channel, approvals waiting on you, actions required, new deliverables or run updates.

Clicking on your organization gives you a complete control panel with 4 tabs: Overview, with all the general information; Members, where you manage your team; Billing, your full financial picture; and last, Usage, a breakdown of consumption for the current cycle.

The capability behind Glob.AI is not built on theory. It reflects what we have been proving on the ground.

In our previous calls, we shared how early AI Pods deployments drove milestone efficiency gains whether accelerating drug discovery research at PharmaMar 15-fold, compressing supply chain contract cycles by 40% at YPF or reducing legacy migration time lines from 14 months down to 2.

What makes Glob.AI so significant today is that those custom high-impact successes are no longer bespoke projects. We have productized those learnings into our standard catalog. Backed by our deep co-engineering alliances with hyperscalers and model providers, Glob.AI turns those proven enterprise outcomes into an on-demand, repeatable capability accessible to every client from day 1. We have built the underlying platform, secured the governance framework and proven the economics at scale. Everything I just described is what compounds behind one number: Glob.AI ARR. We look forward to driving this next chapter together. Thank you very much.

Juan Urthiague

Hello, and good afternoon, everyone. During Q2, we delivered on our revenue guidance, accelerated our AI Pods adoption, launched Glob.AI, grew our top line sequentially and maintained a prudent balance sheet position. We grew on a quarter-over-quarter basis in 5 out of our 8 verticals. And importantly, we grew markedly above company average in our top 50 and top 20 cohorts. Also, in response to observed market volatility, we took actions on our cost structure. I will review our results and then walk you through our updated outlook.

Revenue was $614.4 million, within our guided range, slightly up year-over-year, up 1.2% sequentially. On a year-over-year basis, Q2 revenues included 80 basis points of FX tailwind.

From a geographical standpoint, compared to the prior year period, Europe and Latin America expanded by 6.8% and 5.9%, respectively. Conversely, North America experienced a 2.4% contraction and new markets saw a 17.7% decrease. The new market segment represented a consolidated drag of roughly 115 basis points to the year-over-year revenue growth figure. Due to the ongoing conflict, this specific geography suffered unexpected project delays over the course of the second quarter.

Our cohort performance remains the highlight. Top 50 clients grew 6.9% on a year-over-year basis, top 20 at 6.6%, and top 10 at 4.4%, all well above company average, in line with our 100 Squared strategy. 16 out of our top 20 relationships are showing positive year-over-year growth, and we continue to scale recently signed large deals.

Adjusted gross margin was 36.5%, slightly down as USD weakness accelerated, primarily impacting our largest delivery center, Colombia. And our utilization remained below our targets. Adjusted SG&A accounted for 18.6% of sales, while adjusted operating margin was [ 30.2% ], below our guided range and driven by the impact on margins.

In response to these conditions and to optimally align for subsequent expansion, we initiated a business optimization initiative in Q2. Through this initiative, we ensure the acquisition and retention of the capabilities required for our AI-focused strategy while simultaneously rightsizing our cost baseline to the prevailing market landscape. The main actions under this plan included a comprehensive review of our workforce to align skills and size with our strategic priorities, a consolidation of our global office footprint based on an analysis of our facilities and lease contracts, a strategic prioritization of our delivery centers to support future expansion.

In connection with these actions, we recorded a onetime charge of $32.3 million in the second quarter. And we expect some actions to flow into Q3, which will be critical in protecting our profitability in the short term given the current FX headwinds we are facing, and will be reinvested to fuel our growth engines, specifically our AI platform development and our people.

Despite FX headwinds, we plan to improve margins with the additional efficiencies planned for Q3 and increasing our AI Pods in the mix, which [indiscernible] with margins above company average.

Adjusted net income came in at $60.3 million with a 9.8% adjusted net income margin. Adjusted diluted EPS ended at $1.40.

Our balance sheet remains a source of strength. We ended the quarter with $168.8 million in cash and short-term investments and $253.1 million in net debt. Free cash flow for the quarter was $12.6 million, and free cash flow for the first half of 2026 reached $48.7 million, a record for the company.

On capital allocation, the share repurchase program our Board authorized in May, up to $125 million over 6 quarters, is active. At today's valuation, buying Globant remains one of the highest return investments available to us as the market is pricing Globant as a legacy services company in a soft cycle when what we are is the fastest-scaling AI-native delivery platform in our industry. At the current valuation, the company is trading at double-digit free cash flow yield on a normalized free cash flow basis.

Now let me turn to our outlook. Three external factors have primarily impacted our May expectations, and our revised guidance incorporates all 3. In May, the lower end of our guidance contemplated a significant deterioration in our new markets region that, at the time, was not reflected in our forecast. That scenario materialized, and our expectations for the second half of the year have now changed in the region. Our commitment to the region is long term and important relationships there continue to grow. But the prudent assumption today is that this environment persists in the short term.

Second, we have seen some of the knock-on effects from oil prices, pressuring the travel ecosystem. Some of our travel clients have slowed the pace of their transformation programs to protect their own P&Ls, even as others in the same industry accelerate with us. We believe this is a deferral of ramps, and we expect this revenue to return as industry volatility dissipates.

Finally, we continue to observe protracted cycles in discretionary decision-making. As a result of the above, we are revising our expectations for the second half of the year. For the third quarter of 2026. We now expect revenue to be between $607 million and $615 million. We expect a non-IFRS adjusted operating margin between 13.5% and 14.5%, and the IFRS effective income tax rate in the 21% to 23% range. Adjusted diluted EPS is expected to be between $1.43 and $1.53 per share, assuming an average of 43.2 million diluted shares outstanding.

With respect to the full year, we are revising our revenue guidance to a range of $2.428 billion to $2.462 billion, from $2.462 billion to $2.508 billion previously. In terms of profitability, we now expect our adjusted operating margin for the full year to be between 13.5% and 14.5%, driven by the increasing USD weakness. The IFRS effective income tax rate is expected in the 21% to 23% range. We now expect adjusted diluted EPS of $5.75 to $6.15, assuming 43.6 million average diluted shares.

We expect strong free cash flow generation in the second half, consistent with our seasonality. And our capital allocation priorities are unchanged: the repurchase program and the continued build-out of AI Pods.

The business optimization initiative we carried out this quarter will be visible in our margins as we exit the year, positioning us to enter 2027 with a leaner cost base, record revenue per Glober and the highest margin delivery model, AI Pods, approaching by year-end close to 4% of revenue on a run rate basis.

To conclude, the transition to AI Pods accelerated, we achieved record productivity, and we performed strongly within our top clients. The strong demand we see in AI Pods validates our industry view, one we feel will transform in a positive way. We will be laser-focused on this transition of our delivery model in order to accelerate these trends. Thank you for your continued support.

Arturo Langa

Thank you, Juan, and hi, everyone. [Operator Instructions]

And with that in mind, we will take the first question from the line of Bryan Bergin from TD Cowen.

ช่วงถาม-ตอบ

Bryan Bergin

I wanted to ask on the business transition. So you're showing strong sequential growth in pods, now target of $110 million. I think that's up from $60 million to $100 million before. Based on what you're seeing here, just how long are you anticipating this transition period to be as Glob.AI and the pod model scales whereby it can drive a reacceleration in the overall company trajectory?

And I guess, as it relates to your revised revenue outlook for '26, I think the midpoint of the constant currency forecast is down just under 2%. How much of that is intentional impact as you move under this engagement model versus macro headwinds on the business?

Martín Migoya

Okay. Let me tackle the first one, a very important question. I think that the transition to the new model is something that we are doing it in a deliberate way, and it's something that we will keep on executing quarter-over-quarter. Honestly, the demand that we have seen and the acceptance of [indiscernible] and the positive signs we are seeing from the market are very encouraging. Still is a small percentage, but we think that we will keep on accelerating this.

Now if you ask me, I mean, if the revenue from the new markets wouldn't be affected, we'll be already in the positive growth side without the need of reviewing the whole forecast for the year. So I think it's, overall, it's a very positive movement, and it will accelerate a lot of the growth. Probably by the end of next year, we will see a pretty strong effect of that kicking in.

Now I cannot say right now. I don't think it's a piece of information we can describe now in a very exact way. So with that caveat, I will let the second part to Juan.

Juan Urthiague

Bryan, so the guidance for the year stands now at 2,445 at the midpoint. That is minus 0.4% -- minus -- sorry, minus 0.4%. And the FX tailwind there is about 70 basis points. So on any constant currency, you would be talking about 1.1.

Now when we look at the guidance change, the majority of it is explained by a reduction in the forecast for the new market business. The week after we reported back in May, there was all this news from Saudi, reducing budgets, delaying projects and things like that. And as you know, it's a market that we have been expanding quite nicely over the last few years. And we will continue to do so. We see a lot of deals that are just getting postponed or getting slowed down, but not canceled at all. I mean we keep on having very interesting conversations. So we are confident about the recovery of that market in the near future.

So half of the guidance revision is driven by that. About $10 million is also somehow related to what is happening there because the increase in oil prices impacted some of our businesses in travel and hospitality, and that implied a reduction in the second part of the year forecast for some of those customers.

And then the rest is a little bit of a mix between some assumptions we are doing on certain migrations plus the overall business environment and where we are right now. But again, I think -- and it's important to also look at how the new business and the part of the business that we are pushing very, very hard is evolving. Yes, it is still small. But when you start to compound at 40%, 50% quarter-over-quarter rates, very quick it starts to become more relevant.

And as you pointed out, we had been talking about $60 million to $100 million for this year. Now we are already over $52 million and with very good visibility of the second half of the year. Because we are kind of passing through the first initial stage of trying and testing and understanding what it means to work with an AI Pod. And many of our top customers, actually 45% of the top 20, are already using it. And what that means is that those customers are starting to scale. So we feel confident about the ability to scale this business to over $110 million by the end of the year.

Bryan Bergin

Okay. Just a follow-up here. On the optimization you took, can you just talk about the savings you anticipate from those programs?

Juan Urthiague

Yes. Basically, what we are doing here is we have been reviewing our workforce and aligning that to the current level of demand and also to the current needs of the business with the new models that we have in front of us, and also with the skill sets that are required with a new way of delivering services that we have established. And because of that, we had to make some changes in the organization. Also in terms of delivery centers, we optimized, again, our delivery centers. The impact in the second quarter of that was roughly $32 million. We're expecting around $20 million to $25 million for the third quarter, and that will finalize the program for the year.

We think that that is going to help us first save money, because, otherwise, we would have had part of that talent pool -- or part of that talent -- in the talent pool and without the possibility of allocating them to new projects maybe. And second, it's going to help us offset a massive FX headwind that we are seeing because of the U.S. dollar weakness. I mean if you look at our largest development center, which is Colombia, since the election of the new President, it appreciated almost 15%. And that is a massive, massive impact on our numbers.

So we are going to offset that we are going to invest more. As we were discussing in the call, we just announced a new COO for our -- sorry, Glob.AI business, and we have to invest in that business just because we believe that's the future of the company. So we will be using the money for that, and that will save us costs that we would have had otherwise.

Arturo Langa

The next question comes from the line of Tien-Tsin Huang from JPMorgan.

Tien-Tsin Huang

I think -- I want to ask on the optimization. I just want to make sure I understand just what -- like you said, Martin, you're taking this decision very seriously. What areas were impacted exactly? How much of it was influenced by what you saw in surprising you in May versus the shift to the new model? Or is it really more about the delivery centers and better aligning yourself with some of the FX and inflation trends like you talked to with Colombia. I just want to better understand that.

Martín Migoya

The whole program has like different reasons, right? And you see that on one side, we are migrating to this new model that requires a certain type of Forward Deployed Engineers and certain type of AI engineers that are slightly different from what we used to have, so there's a transition on the talent that we are seeing that is causing one of the reasons of the optimization.

Also, we are seeing like also a transition on the demand of the traditional business. The demand of the traditional business is moving away from web testing into more data and more cloud and implementation. And that transition also created some [indiscernible] new profiles that we didn't have that much before and we needed to start training and retraining. So the effect has that.

Also, it has been -- the impact of several programs and things to run more efficiently the company, as you saw, the increase on the revenue ahead has a lot of -- has a big message inside that as we're becoming more efficient to deliver our revenue. So it's an effort that has like many different components inside of it. I don't know, Juan, if you want to add anything to that?

Juan Urthiague

No. I think -- I mean, it's basically -- it's a real kind of -- not really reshuffle. Maybe reskilling our workforce to the new type of demand, also to the new type of services that we are seeing -- that we are providing to our customers through the airports. You definitely require different skill set.

And also, we need to protect our margins. We need to make sure that we've set all the U.S. dollar weakness that impacts our Latin America business and also make space from the investments that will be required in Glob.AI.

Tien-Tsin Huang

How much was the FX impact in the last year?

Juan Urthiague

If we were to look at all the currencies in Latin America for the last 1.5 years to 2 years, we are talking an overall impact just for the FX of about 4 percentage points. Now we have -- of course, we have been able to increase our revenue per head that help us offset part of that. And we also made some efficiencies last year that also helped us. But the magnitude of the headwind that we have suffered in Latin America has been very, very significant and impacting our margins.

Tien-Tsin Huang

Yes. I appreciate that. And it's out of your control, so it's good to get in front of it. Just on the -- just quickly thinking about the model and Glob.AI, I enjoyed hearing from everybody and Diego. Looking back over Globant's history, I always look at 50 Squared account approach as being really, really important. So thinking about Glob.AI and how this ramps, can you just give us an idea of what the revenue per could be from a client perspective as you penetrate your top 10, 20, 30, as you learn? Is there any analogy or parallel that we can draw back to how Globant under the prior model and assign that to the new model? Just trying to better understand how this can ramp beyond some of the metrics you gave for this year? .

Unknown Executive

Yes. We're seeing very good traction on the 100 Squared program. Indeed, it grew -- that group of customers grew -- the top 50 grew like 7%, something like that. It's extremely encouraging, right? And it is where we're delivering these new things and the first experiences during these last 9 months of execution or a year of execution of our AI Pod. So this is very encouraging by itself.

In terms of amount of revenue, I would say, let's say that we maintain the gross margin as we have a much higher gross margin on the AI bots and we have maybe, let's say, $1.2 to $1 on the new service line. So overall -- but not in all accounts. I mean it depends a lot on every single account. And remember, this is not -- it cannot be traced back to the original model because the original model was headcount, either fixed price or time of materials. This model carries tokens plus token supervision in a single price, either per million token or per output. And that creates a whole -- a totally whole different math, right?

And these 2 models is a place in which you can optimize margins, you can improve supervision. You can do it more with less. Or maybe in some accounts, we need to put more supervision for certain specific projects, but it's managing a totally different manner from before.

And that's why I'd like to say that this transition is not just like a playground that we started. I think it's the future of a company, Moving to output to consumption to value from a totally different model of before. I'm not saying that this old model or traditional model will disappear but I'm saying this is a transition step-by-step, you are seeing us gaining momentum on this new way of delivering, on this new way of charging our customers that is absolutely decoupled from the traditional way.

So making a parallel between those 2 things, sometimes it can seem difficult. And honestly, we have not much story. I mean we have 1 year implementing this. We already have some signals. It's enough for us to put a pricing on those million tokens or this output that we are charging. But we need to see many more things happening to be able to take the kind of conclusion you want.

But what I can say is, listen, we are evolving in a very nice way, growing in a very nice way. It surpasses my own expectations. I said it 60 to 100, now it's at least 110. And I think it will keep on compounding because it makes a lot of sense for our customers. And many of the new projects that I referred to in past earnings calls, about changing interfaces, automating processes, making sense out of massive amount of information, all these things requires a totally different way of delivering.

And this new service is not just for those 100 Squared customers. And this is a beautiful part because this is also to serve other segments which may be smaller, and we are still learning how to do it, but we will keep on expanding on this as we progress. I don't know, Juan or Fer, if you want to add something?

Fernando Matzkin

No. Like the last thing you said, it's quite important with Glob.AI, we're also thinking of how to widen our base of clients, how to serve clients with different kind of scale that we couldn't do it before in different segments to accelerate our revenues. We've proven the success of the 100 Squared model, like you said, and it is very well reflected on the growth of at least our 50 top customers. And now the challenges like sustaining the growth of a segment. Our clients have different dynamics, right, and needs to be so different. And we believe that with Glob.AI, we also have a way to reach a much wider base of customers with a simpler way and a more sustainable way.

Arturo Langa

The next question comes from the line of Maggie Nolan from William Blair.

Margaret Nolan

Maybe I wanted to build on one of those past questions. You -- let's see. Okay. Yes. So you had talked about in the comments that you thought that Globant AI and having a focus on outcomes was opening up new budgets to you that you hadn't had access to before. So maybe can you elaborate on where that growth is coming from, who are the new buyers? Is that growing? Or do you view a total addressable market as growing? And kind of what's changing your ability maybe to go deeper in clients?

Unknown Executive

Well, look, that paragraph that you mentioned refers to the following. For years, as I described at the very beginning of the earnings call, we have been meeting experiences and software products and we have been very close to using technology to create experiences that engage in an emotional way with our customers. And that is the core. That's what's -- define us.

Now as AI came, there are many other places in which it's not just creating those experience, but also operating part of those back-ends and processes that before were not an opportunity for us. So as I described that the process automation and the change on the landscape of how do you use AI for pretty much everything, including automating process, creating new or charts that reflect that automation, we believe that our AI Pods, which are now the AI Pod software that what you see in Glob.AI if you go will evolve into operations and will include that same concept of having an agent operating something for you and having humans being able to analyze the edge cases and charging in a way that is per unit of that specific business case, let's say, travel. Well, it will be per trip or Know Your Customer, per Know Your Customer.

So that agentic work that this Glob.AI concept is opening up is much larger than the software development life cycle that we have right now on the Glob.AI. So what I'm saying is, with this new idea with this new concept, with this new definition of how services and AI native services will be rendered, we are able to tackle much more than just the original software development life cycle, and we can expand our presence into AI Pods for operations, right? So that's what I'm referring about when tackling new budgets.

And also there's a market share gain too, because when you present this new way of doing things and new way of charging things, for every dollar you sell on this new model, you are able to capture maybe another do, right, of that same budget because you are doing things that other vendors we are doing. So we're extremely excited about those 2 things happening at the same time, expanding into other places and capturing more dollars for that work that we used to do.

And in many cases, it was the reason why we win and we won. I have many examples, several examples otherwise we would have lost. And I think that this new definition of how to do things and how to charge for things is a really new approach to our customers. And we launched it last Thursday. And the impact, the amount of people that call us, the amount of people that are interested in understanding more about how this Glob.AI model works, in essence, it's a huge effort and a huge intellectual effort in creating something that didn't exist. And we are the first providing it.

So I'm extremely proud about the whole team that is developing this. I'm extremely proud that we have been able to take it to the market, and we are being able to convince our customers. And not just a couple of cases, now it will be more than $100 million in ARR. That is an absolute success. And I think that this is something that you will see us insisting more and more. And deliberately, enter into -- asking our customers to change the model. And that will be a process. That will be a process itself. It reminds me to some other companies changing how they do business. But I think overall, it will be very exciting to see that transition and to run it.

Margaret Nolan

Martin, that's super helpful. Maybe then obviously, AI Pods is the growth driver here and Globant AI. But EMEA was expected to be kind of a future growth driver for the company and has been an important region in the past couple of years. So you were pretty clear that you were conservative or prudent in your outlook for that region as the dynamics there have changed.

I'm wondering where, from an end market perspective, whether it be vertical or geography, you're turning your attention to as a potential growth driver over the next kind of 12 to 18 months, and what we should look for, for success metrics there.

Martín Migoya

Excellent question, Maggie. Listen, the ocean of our industry is pretty flat. I mean you see other companies, they are pretty much all of us in the same kind of level of growth. What we found is that is not just the old model with new tools, the success, but a totally different delivery system. When I see that we found a place that is growing fast, and we want to build our company around it, is exactly what we want to do.

Now there are some industries that are taking this faster than others. We are seeing a lot of success in financial services, in media and entertainment. We're seeing a lot of success in airlines. This thing make things more efficiently and faster, 30% faster. So as we see the game evolving from cost reduction that has been the main focus of everybody with AI today to revenue generation, which is what I believe is the smart way of using AI, then we will see a lot of industries coming into this space.

So the message I want to convey is not just a region or an industry, but it's a new way of delivering what you should pay attention to. And that's why I ask you to hold us accountable to that number of transition as we evolve this company. This is where we are putting our energy, on how to deliver in a much more efficient way. It's not just adding people plus AI. People plus AI means a lot of slop, a lot of time used to supervise that slop, a lot of rework over and over and over. And when you put order in that process like what we do with our Glob.AI way of delivering, then things become much more efficient and things become independent on the model that you want to use and things become scalable.

And then Glob.AI is just -- this is just the beginning of Glob.AI. Glob.AI is much broader than that. Glob.AI is -- the initiative we will use to transform Globant into an AI-native company, the whole Globant into an AI-native company. And it is our AI native services arm. So let's see, it's not just -- I cannot answer with just space, a region or an industry. This is wide; this is very wide.

Arturo Langa

The next question comes from the line of Bryan Keane from Citi.

Bryan Keane

Just wanted to ask, Martin, when you talked about you're making a choice to push more work to AI Pods, and that seems like it's costing your existing business or hurts the core revenue of that business, can you just talk about that deflationary pressure and why that doesn't last long as we go for this transition over the next couple of years? Are we going to have to kind of run in this negative revenue territory due to the deflationary pressure that may be pushing work to the AI Pod model is going to cause?

Martín Migoya

Well, look, I mean, in essence, a lot of our customers have been spending the same amount of money getting more productivity, right? And that has been the case in the vast majority of the things. I think in the future, that could evolve to, first, we need to convince procurement, we need to convince more people, and that process could be slower than just running the traditional game. But we're ready to play that. And I think that the margin overall will be much better. And the capabilities for us to improve the margins even further from where we are today is still better.

So I think that that deflationary scenario is something that we're not seeing right now. If we need to pay for it, we will pay it, I want to do that transition. And that's why I said, listen, we are migrating this even understanding that in some accounts, this will be some demand -- softness in demand -- soften demand, but I believe that overall, the picture will be totally different. I don't know...

Juan Urthiague

I was going to say, Bryan, that sometimes what we've seen is that it may take longer, a lot longer, to convince to persuade the procurement teams and our customer to transition and keeping the same level of revenues with more productivity. So sometimes, we're seeing that -- we know -- we want to migrate, okay? And sometimes, we can do it faster if we're willing to provide some efficiencies immediately to the customer in terms of price. But we believe that because the model is so much more efficient and so it makes so much more sense for them, that we should be able either to expand on other areas of the organization, to win market share from other vendors because we -- I mean what we believe is that this is a model in which we need to deliver services. Now if we can do it faster, we will do it. And I think that's a key message that we are giving here. .

Bryan Keane

No, that's really helpful. And then just as a follow-up, Juan, the revenue per head, it jumped to almost double digits. How much of that is like-for-like pricing? Or are you guys getting a little better pricing right now in the market? Just trying to understand that number.

Juan Urthiague

It's a combination of different things. The market is competitive, right? There are some occasions where we are being able to deliver with less headcount because we are being more efficient with our delivery model right now. And if you look at the total headcount, it's down roughly almost 10% year-over-year with revenue per head going up almost 10% year-over-year. So we are being more efficient.

In some cases, we have been able to get some additional pricing. But it's -- I mean, I wouldn't take that as the norm because the market is very competitive right now. But we have been able to increase our revenue per employee because we are delivering in a more efficient manner.

In some cases, we are charging that with the new model, with the AI Pod revenue or the Glob.AI revenue model. In other cases, it may be a fixed price where we are able to deliver more efficiently and, hence, increasing the revenue per car. So it's not that we are charging like Glob.AI, because it may be a fixed price, but we are still getting more revenue per employee because we are delivering more efficiently.

And I think that explains the sharp increase. If you look at the revenue per head, 3, 4, 5 years ago, it was around $60,000, $65,000 per employee. Now we're getting close to over $90,000 and getting close to $100,000.

Arturo Langa

The next question comes from the line of Arvind Ramnani from Truist Securities.

Arvind Ramnani

Just had a couple of questions on Glob.AI. Martin, you said that's kind of where you're focused your efforts on. So just a couple of questions out there, right? Like how does the workload split across these OpenAI and Anthropic and open weight models today? And how do you expect that to shift over the next like 18 months?

And then second question around that is you also mentioned a lot of those clients who are leveraging Glob.AI is existing clients. But how many clients -- or what percentage are like kind of new clients, right? Like who are not using that?

And just last question on that is that what's kind of Globant's kind of unique value proposition, right? Is it around enterprise context? Is it around routing logic? Like what's proprietary to you?

Martín Migoya

Okay. Let me -- I will start with the first and then -- the last, and I will let Diego to complete. I see that we just announced a partnership with Anthropic. We are extremely excited with the things we can do together. We are already seeing some impact on -- from that pipeline coming into our scope of work, which is very, very exciting. I see models have -- models will be evolving. And our customers will decide kind of what to use. When we see Glob.AI, we see something that it could use pretty much any model on the core when developing the software or creating the software, including open weight models, if that's the case.

We're processing a big chunk of our tokens with our own infrastructure and our own models, using open weight models in many occasions. And some of our customers are requesting that. Some of our customers are saying "We'll go full-fledged with Anthropic or with OpenAI." So we have pretty much full independence on that specific thing. And of course, tools will keep on evolving and keep on becoming more and more sophisticated.

Two, the specific moat that I would like to describe about Glob.AI, Glob.AI is a play of services. Basically, we are mixing in an absolute frictionless way. The creation of the experience and the creation of the software and the definitions that you need with AI and with humans and putting -- and packing everything into a single price and a single price per consumption or per output.

So basically, what you saw, the demo, the video that Diego showed to us, it's a video that makes it very clear that Glob.AI helps you with the definition, helps you with the creation of the specification, and then a set of agents get triggered, a set of loops get triggered or workflow gets triggered Depending on what you need to do, those things are different, and these playlists have been graded and evolve with time. And depending on which player you are using they were required like different levels of supervision for humans that are watching what those patients are creating, but we charge you in an extremely transparent way, either per million token or per output in case of a user story or in case of -- so what I'm saying is that this coordination, that elimination of the friction to buy services is the real thing that we are providing. And that's independent on any model that you may choose, right?

So that's what we are saying. This is real AI-native service. Because it's playing the same role that in the past when cloud didn't exist and we needed to create compute, you buy the servers, you buy the connectivity, you used to buy the -- or to hire the people to manage those servers. And then suddenly someone coordinated everything in a beautiful way and it was so easy to ramp up infrastructure, right?

Well, professional services is in that old era. And with Glob.AI, we're creating the AWS of the services, right? So it's extremely easy to go explain your project, create the context, connect with your JIRA or with your GitHub or with your -- whatever repository you want, and then execute the mission and that mission will be played in an extremely professional manner, supervised by the best people that can lead you to the right enterprise results you are looking for.

That concept, it seems too easy, but it's extremely sophisticated. It's the evolution of services. And it's not just a platform. And it can play with Codex, with our own CODA. It can play with Claude. It can play with pretty much any of the AI tools that are out there. And then it can be played into any type of infrastructure. But the thing is it always gets coordinated in a pretty efficient manner with a human. So no more ramp-up of teams, no more long time of procurement for something. No more not understanding how much something will cost. It's a totally different thing. And that's why Diego showed the video. But I don't want to screw your speech.

Diego Tartara

Sorry, I think it's good and it's a totally valid question. One of the things that we do find every single [indiscernible] project has a ton to do with accountability and repeatability. You will not get accountability from a frontier model. You use it, you get a result. Is it good? Is it bad? It's up to you, you implement? That's what you get. Globant brings you accountability. That's why we have humans.

And the second most important aspect is that the model is not repeatable. When you go to a frontier model, it does -- I mean, generating out is super good. It's amazing. And we are capturing 100% of the value there. But when it comes to planning, how do I execute something, it's a combination of how you prompted, how much information you gave it, what's the decision and thought process on the model. It consumes a ton of tokens for solving sometimes easy tasks that could have been solved with probably 10% of the spend, as an example and we actually moved away from that. And [indiscernible] are actually the formula of how the proper -- what's the proper way of delivering this type of value. What's the proper way of doing a migration, a replatforming, SAP4HANA migration, as an example?

The blueprint, it has a ton of very well-defined steps. Every step it has a required input, not less, not more than that, in a required output, and a supervision for that, which is called a quality [indiscernible]. So you get a repeatable system, a system that you can feed over and over and you get the same. This is something you don't get out of AI and drives enterprises crazy.

So you don't have control on the spend in many occasions. Models are actually working with both ends, the definition and the testing. And what happens in the middle kind of looks like brute force, like trial and error, and that's how graph engineering in many occasions work. So we totally changed this, and we use what we know, where the humans can actually add value, make sure things are completely right, and you provide the accountability for that.

So coming back to your original question, I think that every single company we talk to actually finds a lot of comfort and feel a lot more comfortable with this model. This is the type of services we've been providing for over 20 years. It's not about the output. It's not the source code. It's not the executable. It's about the process, capturing what the client actually needs, providing value on top of that, and holding yourself accountable for that output, for that business impact.

Martín Migoya

And let me illustrate this with one example, which is from a few days ago. And what happened was one of our customers at Glob.AI to some kind of architecture definition for a pretty complex project around ERPs and APIs and connections. And he dropped it on Glob.AI, our guys guided, I mean, the agents start to do the work. Supervision happens across that work. We interact with the customer 2 or 3 times. And we finished that in a record 48 hours.

Now to do that same thing, even with Globant in a traditional way, would have taken at least 2 or 3 weeks, because of the meetings and the things and the scope gathering, all those things, all that process that was extremely inefficient before has been concentrated in a very simple way of doing it now.

So that value that is created by understanding the customer faster and getting to the accountable result, as Diego was describing, is the main thing that we are talking about today. And that's the transformation I want to make for the whole company, for the whole Globant and for our whole customers. And I think that yields much better margins, much more predictable revenue, much more consistent and recurring revenue. It yields, I believe, results for our customers that are way beyond just using AI with a set of engineers. And I think it's the answer that many will follow. And it's not just us. I mean, we are starters and we are innovators in this mission. But it won't stop here. So it will be fun.

Arturo Langa

The next question comes from the line of Jonathan Lee from Guggenheim.

Yu Lee

Great. Good to see you guys. Martin, I appreciate the vision, and I understand you're not seeing deflationary pressure today around AI Pod work. But how are you thinking about combating it and when it does emerge, particularly when clients come back on renewal and demand a larger share of productivity gains? What are the structural defenses in the pod model that led you hold pricing when procurement inevitably pushes back?

Martín Migoya

I didn't get the question exactly, please. Can you repeat it again?

Yu Lee

So as you think about the deflationary pressure that you're not seeing today, what happens when you need to combat it going forward if it does emerge? And is there anything structural in the pod model that lets you hold pricing when procurement pushes back on pricing?

Martín Migoya

Well, okay. Listen, I think that if procurement comes back, that's always a negotiation, but that happens in every single model, not just in the AI Pod. But the thing is I believe that we can be much more efficient. If the customer wants to do the same, it will cost less money. If the customer wants to do more, which is most of the cases that we are finding, they will spend the same amount of money. And the thing is that the customers will want to do a lot more. So I think there will be an up-spending in this new model, more money to produce much more as we have been describing in the past.

So again, this is a game about the amount of software and the amount of solutions and processes that must be created or used in this new era. So I think the game cannot be predicted that simple saying -- putting like a constant value on the amount of things and [indiscernible] that can be do -- or that are needed, sorry. So as this amount of things moves everywhere, and in many occasions, moves up as the new needs, as described before, the new needs happen, then where you land with the AI Pods there, I think, in my opinion, will be increasing.

Now that's demonstrated on the top accounts that are growing at 6% and 6.9%.

Juan Urthiague

And also even in that scenario, potentially, when a customer wants to do the same and there is a potential saving for the customer there, the model runs at higher margins than the traditional models. And maybe we will be making the same dollar amount with a little bit less revenue on some accounts. But at the end of the day, the way we look at this is, look, this is the way services need to be delivered going forward.

If we do it better and faster than others, even if we lose some money in some projects where we need to reduce a little bit of the revenue, we will earn more projects, we will win market share and we will eventually grow faster again. So that's how we are looking at it. And we are willing to take some of those cases. Again, it's not in every case. We've seen many cases so far where it just keeps adding more work. But there are some occasions that we are seeing that if we take a haircut, we can accelerate the migration.

And this is what we are saying today. We are willing to accelerate because it gives us a better position in front of the customer. It protects us from the competition. And we believe that it's going to drive more business into Globant going forward. And at the same time, because we can be more efficient, especially as we scale, there is more margin to be along the way.

So there are multiple things happening at the same time. It's still hard to move [indiscernible] for us. We are building models every day. But we are seeing at least that there is clear traction, that there is a clear improvement in margins, and the customers are enjoying and are coming back to scale the model. So that's what we are looking at this.

Martín Migoya

Yes. And the game and the play for us is we found that space that is growing fast. We want to expand that transformation and do it faster. This is a very simple way of putting it.

Diego Tartara

And I think one additional thing is that, as we all know, the market as it is today has been concentrating on the cost saving machine and operational side of things. And it's a lot more difficult, I think it's a worse scenario because it's a lot more difficult when you're part of the cost equation, right? And we -- and that's the type of pressure and companies may want to reclaim part of those efficiencies.

Once the markets start recovering and moving to a revenue-generating engine and stop neglecting products, as an example, things will definitely change. We've seen this over and over. When the target -- the product, you want to squeeze in new features, additional content, et cetera, so it's not about saving the money, it's about making your product better and stand out in front of others. So hopefully, we will see that happening soon.

Yu Lee

And just as a follow-up, what macro backdrop is assumed in the revised '26 outlook? I mean does that range assume conditions persist maybe some stabilization in North American decision cycles or continued deterioration? And how much cushion is built into the low end for further softening?

Juan Urthiague

No, I think that the most likely scenario is the midpoint, but it basically assumes already a significant impact from the Middle East and our travel customers. Of course, I mean, again, if things get a little bit better on the macro side, especially in the U.S., we can probably be a little bit above that range. And again, if things get worse, we are not seeing that now, but the cushion is there just to make sure that the range is something we will achieve no matter what.

But the most likely scenario for us, as always, is the midpoint. And unfortunately, we have seen some factors that impacted our previous guidance that we could not control. But the midpoint is our most likely scenario. We have plans to get higher than that. We will try to execute on those. There are some things that we don't control.

Unknown Executive

And to add to that, 90% of our revenues throughout the year are already contracted. And in some of the markets that have been affected, like particularly in new markets, we have pivoted the pipeline towards banking, towards public sector, areas that are less affected as like entertainment and travel and hospitality, which was where our focus was put in new markets so far. So we are rebuilding the pipeline. We are closing some of those new opportunities. So we also expect that new market suffered quite a lot. This quarter, we also expect a quick recovery towards the end of the year.

Arturo Langa

So that will be off for the Q&A section today. Thank you all for your time. And now I will ask Martin to provide some closing comments.

Martín Migoya

Thank you, Arturo. Thank you, everybody, for being here today. I'm looking forward -- and thank you for your continued support. I'm looking forward to see you on our next quarter call. Thank you. Goodbye.