Tesla Q4 2025 Preview: When Cars Slow, Can the AI Story Carry the Stock?

Over the past year, most media discussion around Tesla has revolved around a single theme: falling deliveries, price wars, and margin pressure, as if any shortfall in units shipped means the company is finished. But if you’re still only watching the auto business this quarter, you’re already looking at the wrong battlefield; what really drives the valuation now is FSD subscriptions, Robotaxis, Optimus, and a whole AI and compute network that is starting to take shape.

You can think of Tesla as a potential compute map: millions of cars already equipped with the latest-generation self‑driving computers, spending most of their time parked, each with built‑in power, connectivity, and cooling. The story Musk is telling now is no longer just about cars that can drive themselves and pick up passengers, but about whether these idle in‑car computers can one day become distributed AI nodes that power AI applications—and then share part of the compute revenue back with the owners.

From this angle, the Q4 2025 delivery, profit, and inventory numbers still matter, but something else matters more: does FSD start to throw off meaningful subscription cash flow, do Robotaxis and Optimus move from slide decks into real operations, and does the puzzle of energy plus fleet compute fit together one step further into a coherent network?

FSD’s Lifetime Option Ending: From One‑off Sales to A SaaS Business

The next change is crucial: Tesla has officially announced that after 14 February 2026, it will no longer offer FSD as a one‑time lifetime purchase and will switch fully to monthly subscriptions; the current window for free FSD transfers also closes on 31 March. In the US, FSD currently costs 99 dollars a month.

Why does this matter? Because this is Tesla’s biggest reset in accounting and business model since the Model 3 scaled up. Previously, it was about taking 8,000 dollars up front; going forward, it’s 99 dollars per month. In the past, you were essentially buying a lifetime ticket to “self‑driving someday”; now it looks much more like paying only for the capabilities you have today, with future upgrades explicitly sold as ongoing services.

In the near term, you might see a bit of revenue deferred out of Q4 2025 and Q1 2026, as less of it can be recognized up front and more will be spread over the coming years. But for valuation, the meaning is completely different: this starts to look like a SaaS company, where the focus shifts to ARPU, churn, subscription growth, and whether Robotaxi services can be layered on top.

That’s why, on this Q4 earnings call, the market will be watching closely to see whether management starts giving FSD metrics the way a subscription business would: what share of new cars opt for FSD subscriptions, what is average monthly spend per user, what is monthly churn, and how heavily is FSD actually used each day. If these metrics gradually surface over the next few quarters, you end up with two completely different valuation stories: a carmaker versus a software‑and‑compute platform layered on top of a fleet.

Robotaxi: From YouTube Demo to A Line Item in the P&L

Back to Robotaxis. In Texas, Tesla has already obtained a Transportation Network Company (TNC) license, valid through 6 August 2026, allowing it to operate ride‑hailing across the state, with rules that permit both supervised and fully driverless service. Since June 2025, Austin has seen small‑scale operations begin with a safety driver in the front seat, holding a kill‑switch.

Musk once boasted that “Austin will have driverless cars by year‑end,” but that milestone has clearly slipped: only in late January this year did Tesla start putting a small number of truly safety‑driver‑free Robotaxis on the road in Austin. Even now, the fleet is a mix of a minority of unsupervised vehicles and a majority still under human monitoring, with some cars even tailed by chase vehicles—human oversight is far from gone.

If from this Q4 or the next few quarters Tesla is willing to disclose hard numbers for Robotaxi in Austin—daily trips, utilization, cost per mile, pricing—then FSD can evolve from a YouTube demo into a business unit with its own line in the P&L. But if Robotaxis remain stuck at the storytelling stage, while regulatory approvals in Europe and China continue to slip (the market’s optimistic case is for initial decisions around February 2026), then the discount applied to this part of the story will only grow, weighing on the stock.

Optimus + Macrohard: One Does Physical Labor, the Other Office Work

Having covered FSD in the car, it’s time for the robots. The 2025 roadmap here is already fairly clear: Fremont now has a pilot production line for Optimus, with theoretical annual capacity of up to 1 million units; at Giga Texas, Tesla is planning a dedicated Optimus plant with 10 million units of annual capacity, with site clearing and utility work under way, major construction expected in 2026, and volume production in 2027. Before that, Tesla intends to deploy small numbers inside its own factories, putting Optimus onto the most monotonous and repetitive stations, with a goal of reaching several thousand internal units around 2026 to validate at scale.

You can think of the physical Optimus as a human stand‑in for the real world: any physical job a human can do, in principle it can take over, at a much lower unit cost, running 24/7, no sick leave and no slacking. xAI’s Macrohard digital worker project, by contrast, is a human stand‑in for the digital world: any task that needs a keyboard, mouse, screen, and decisions, it can mimic a human to complete.

Underneath both, the same technology base is shared: a unified neural world simulator. Tesla uses fleet video to train a world model that can drive cars in the physical world (FSD), guide Optimus through factory work, and also serve as the “brain” of a human simulator.

If, on the Q4 call, Tesla can offer even a sliver of hard data—for example, how many Optimus units a given factory has deployed, how many human workers they replaced, and the per‑hour cost advantage versus labor—plus capex details for the 10‑million‑unit Texas plant, then the post‑2027 cash‑flow story will take a big step away from sci‑fi and towards engineering reality.

From Colossus to Megapack: The Energy Business Is Really Fighting the AI Infrastructure Battle

On the Colossus data center side, to stand up a full supercomputer and power system in 122 days, Tesla reportedly pulled in over 80 mobile generators plus large battery packs to smooth multi‑megawatt load swings. With GPUs ramping power up and down, if you don’t do local storage and just slam it straight into the grid, you genuinely risk destabilizing local networks.

That sounds dramatic, but from Tesla’s perspective it’s a very direct commercial opportunity. The problems Colossus is facing are the same ones every AI data center in the world will face; the solution, conveniently, is exactly what Megapack is built for. And in storage, Tesla has put up serious numbers: in the latest reported quarter, deployments hit 14.2 GWh and 46.7 GWh for the full year, up nearly 50% year‑on‑year, with gross margins around 30%; sustaining a 20–25% margin over the long run looks entirely plausible and is already clearly higher than autos.finance.

By the numbers, the market is looking for roughly 3.8 billion dollars of Q4 revenue from Tesla’s energy business, about 12.75 billion for full‑year 2025, and potentially close to 18 billion in 2026. This quarter, energy is no longer just a margin buffer for autos; it’s the arms dealer in Musk’s AI war—supplying xAI today, and potentially selling Megapacks to Anthropic, OpenAI, and other rivals tomorrow.

Q4 Headline Expectations

After all this story‑telling, we still need to land on the numbers. Here’s where consensus currently sits for Q4 and full‑year 2025:

On the auto side, not pretty is already consensus. Full‑year 2025 deliveries of 1.636 million vehicles are down 8.6% year‑on‑year, marking a second straight annual decline. Q4 alone saw 418,227 deliveries, about 1.1% below consensus and 15.6% lower than a year earlier. In Europe, annual volumes fell by close to 28%, while in China, Tesla faces a wall of domestic competitors led by BYD.

Meanwhile, capex is surging. Tesla’s 2025 capital spending is around 9 billion dollars, and management has already signaled a significant step‑up in 2026, largely aimed at AI infrastructure: FSD training compute, Optimus production lines, and the factories and datacenters needed to support Robotaxis and large‑scale compute clusters.

That puts a key question squarely in front of investors: of the pressure you see today in gross margins, how much is pure damage from price wars, and how much is investment pain from AI spending? If the latter is a big chunk, are investors willing to treat that margin compression as upfront capex for future FSD, Robotaxi, Optimus, and compute‑platform cash flows? That’s the story this Q4 call has to make clear.

In the End: Q4 Is Less About the Print and More About the Narrative Turn

For many long‑term holders, Tesla’s appeal has never just been a string of financial ratios. It’s about a sense of participation along the time axis: watching a company that once “just made EVs” steadily drag things like self‑driving, humanoid robots, and grid‑scale storage out of science fiction and into reality, piece by piece. Musk’s real gift has never been getting timelines exactly right; it’s that, every so often, he actually drags one of those imagined scenes down onto real roads, factory floors, or power grids. Sometimes he’s early, often he’s very late, but there’s always that moment when you suddenly realize: this thing is actually running.

Over this year, the focus will be on this checklist:

This quarter, what many people really want is a blunt answer: after two straight years of falling deliveries, a bruising price war, and surging capex, is the long‑dated story around FSD, Robotaxis, Optimus, and energy still one the market is willing to underwrite? The numbers will tell you how much Tesla is earning right now; whether those boxes—FSD subscriptions, Robotaxi operations, robots, and Megapacks—start filling in with real data will decide whether, in investors’ minds, Tesla is just another automaker to be rerated on traditional metrics or still a portfolio of long‑duration narratives worth valuing on a different curve. The rest will come down to the moment after the close on 28 January—whether, and to what extent, people on the other side of the screen are still willing to commit their time and capital to the next chapter of this story.