Business

Know the Business

Bottom line: Tesla is no longer a pure-play EV manufacturer — it's a capital-heavy vertically-integrated hardware company whose equity is priced almost entirely on three optionalities that don't yet generate meaningful cash: FSD/Robotaxi, Megapack-scale energy storage, and Optimus. The core auto business is shrinking (revenue −3% in FY2025), auto gross margin has halved from the 2022 peak, and the ~$2B/year regulatory-credit income stream is being curtailed by the OBBBA. The market is paying 371× earnings for a business where the operating engine and the valuation engine are two different companies — and it's the smaller, unproven one being paid for.

1. How This Business Actually Works

Takeaway: Tesla runs an auto OEM (shrinking, thinning margins) that funds two growth bets — energy storage (scaling, margin-expanding) and AI/autonomy (pre-revenue at scale). Incremental dollars of profit no longer come from selling another car.

Three streams tell the real operating story. Auto sales ($65.8B, −9%) shrank on both volume (~8% fewer deliveries) and mix/price (higher incentives, attractive leasing to move inventory). Services & other (+19%, now $12.5B) is the emerging annuity — paid Supercharging, insurance, used-vehicle flow and non-warranty service — and carries near-breakeven gross margin today with room to widen as Supercharging scales. Energy generation & storage (+27%, now $12.8B) deployed 46.7 GWh of Megapack/Powerwall and expanded gross margin from 26.2% to 29.8% — this is the only segment where both revenue and unit economics are moving the right way.

The two streams that flattered prior-year earnings are both eroding. Regulatory credits fell 28% to $1.99B as the OBBBA repealed consumer EV tax credits and curtailed credit programs tied to Tesla's products — this revenue has near-100% incremental margin, so its decay hits operating income disproportionately. Leasing is down 6%.

Where the incremental profit comes from now: software (FSD subscriptions, in-app upgrades, premium connectivity — recognized over the life of the vehicle), Supercharging revenue from a captive fleet, energy storage where utility-scale demand is AI-data-center-driven, and — if it works — Robotaxi ride fees and Optimus. None of these is capital-light on day one. Capex is guided above $20B in 2026 (vs $8.5B in 2025), driven by AI compute, Cortex 2 training clusters and new vehicle/cell lines. The economic engine has become: use scaled auto operating cash (~$14.7B OCF) to fund AI/robotics infrastructure that is priced as the real franchise.

2. The Playing Field

Takeaway: On scale, Tesla is a mid-sized OEM (half of Ford/GM by revenue, one-tenth of BYD by unit volume). On profitability, it is the highest-margin listed US auto-maker in FY2025 — but the gap versus legacy OEMs has compressed dramatically, and against BYD, Tesla's margin and growth advantages are now gone.

BYD figures are approximate from latest Hong Kong disclosures; Stellantis FY2025 reflects ~$13B impairment/restructuring charges that drove the GAAP loss — normalized GM is ~11%.

The peer matrix reveals the central tension. Tesla trades at ~20× the market cap of GM on smaller revenue and a thinner operating margin than Ford or GM at their peaks. The bull case has always been: Tesla is not an auto peer — it is an AI/software/robotics platform. The bear case is: at today's margin structure (4.6% operating, 17.8% auto gross), Tesla looks a lot more like a premium auto OEM than a software franchise.

The two peers that matter most are BYD and legacy GM/Ford. BYD now out-produces Tesla in EVs, operates at comparable gross margin with a faster-growing top line, and has its own battery vertical — it is the first real head-to-head competitor Tesla has ever faced. Ford and GM have the opposite problem: they run large ICE businesses that still print cash, while their EV programs lose money. The startups (Rivian, Lucid) remain sub-scale cash consumers — useful only as a reminder of how expensive it is to stand up EV manufacturing from zero.

3. Is This Business Cyclical?

Takeaway: Yes, but not in the way legacy auto is cyclical. Tesla's revenue hasn't contracted meaningfully outside the 2024 ASP-led compression — the cycle hits through margins, not volumes, because pricing is the lever Tesla pulls to protect utilization at its gigafactories.

The 2022 peak (16.8% op margin on $81B revenue) was a once-in-a-cycle combination of: post-COVID demand overhang, constrained EV supply globally, high ASPs on Model Y before Chinese competition scaled, and a surge of regulatory credit buyers. From there, margin compressed every year — pricing cuts in 2023 to defend volume, incentives and inventory charges in 2024, and in 2025 a combination of tariff pass-through, lower fixed-cost absorption from weaker deliveries, a 28% drop in regulatory credits, and an 41% R&D increase tied to AI build-out.

What the cycle looks like for Tesla: unit volume is fairly resilient (deliveries held around 1.8M in 2023–2024 and only dipped to ~1.64M in 2025), but ASP and mix move sharply. Gross margin is the stress variable — it compressed from 25.6% in 2022 to 18.0% in 2025 even though revenue barely moved. Capex is pro-cyclical on the way up (gigafactories, Cortex) and stays elevated in downturns because the investments are AI/compute rather than capacity-driven; 2026 capex is guided above $20B against FY2025 OCF of $14.7B — implying Tesla may be a net cash user in 2026 for the first time since 2019 if deliveries don't inflect.

4. The Metrics That Actually Matter

Takeaway: Ignore reported EPS for the next 24 months — OBBBA credit changes, bitcoin mark-to-market, stock-based comp, and AI-related restructuring charges make the P&L noisy. Watch these five.

2026 OCF shown as flat placeholder; 2026 capex per company guidance ">$20B". The bar chart flags the crossover, not a forecast.

Why standard ratios mislead here. P/E of 371× reflects an assumption that AI-driven earnings will re-rate the denominator by an order of magnitude — if you were evaluating a pure auto OEM you'd laugh at the multiple, but nobody owns Tesla at this valuation for the auto P&L. ROE of 4.6% is similarly deceptive: it's been compressed by the balance-sheet swelling (cash/investments grew from $36.5B to $44B) rather than earnings collapse alone. Debt/EBITDA is essentially zero; the company has no leverage question to answer.

5. What I'd Tell a Young Analyst

Takeaway: Build your model on the parts you can see, and size the parts you can't. Don't anchor to either the bull case ($3T+ AI platform) or the bear case (another Stellantis). The real work is triangulating how much of the current market cap is auto + energy + cash, and therefore how much is implicitly paid for FSD/Robotaxi/Optimus — then stress-testing what each of those needs to deliver to be worth that residual.

Four things to carry into any Tesla call:

1. The regulatory credit line is not an accounting curiosity — it is 100% of FY2025 operating income ex-energy. $1.99B of credits vs $4.36B reported operating income means roughly half of reported profit is ~100%-margin revenue that the OBBBA is phasing out. Strip it out and FY25 operating margin is closer to 2.5% — a margin structure that does not justify a premium auto multiple, let alone a 371× P/E.

2. Energy is the most underappreciated piece of the story. Segment revenue grew 27% to $12.8B with gross margin expanding 360 bps to 29.8%. Megapack demand is tied to AI-data-center grid stabilization, not consumer sentiment or EV subsidies. If you want a defensible, margin-expanding, scaling franchise inside Tesla today, this is it — but nobody talks about it because it's not the shiny object.

3. Capex is inflecting from 'manufacturing' to 'AI infrastructure.' 2026 guidance of $20B+ capex is the tell — this is Cortex 2, data centers, and compute, not another gigafactory. The company is making a binary bet that real-world AI compute trained on fleet data is a durable moat. If that bet works, the valuation is cheap. If it doesn't, the capex compounds into a second decade of suppressed FCF.

4. Watch the auto GM trendline and the FSD attach rate — both are real-time confidence votes on the thesis. Auto GM ex-credits is the hard number for the core product; FSD subscription revenue ramp is the hard number for the optionality. If auto GM ex-credits keeps sliding and FSD revenue doesn't visibly inflect in 2026, the thesis is cracking no matter what management says about Optimus.

What would genuinely change my mind (either direction):

• Bullish trigger: Robotaxi per-vehicle-day revenue disclosed at levels approaching $200+/day at a commercial scale — that would validate a fleet-based software business worth the premium.
• Bearish trigger: Energy storage margin reversal below 20% plus sustained auto-delivery declines into 2026 — together, that would mean the two healthiest vectors of the business are both breaking.

The rest is noise.