$725 Billion Is About to Get Audited: What Alphabet, Tesla and Intel's Earnings Actually Test

Alphabet and Tesla report after the US close today. Intel follows tomorrow. Together the three are worth well over $5 trillion, and the framing that has settled over this week is that these results are the first real test of AI capital expenditure at scale.

That framing is half right. Two quarters of results cannot settle whether roughly $725 billion of 2026 hyperscaler capital spending will earn its cost of capital — that question resolves over years, not on an earnings call. But the numbers reported in the next 48 hours will tell you something narrower and more useful: whether the companies doing the spending are still willing to defend it, and whether the market is still willing to accept the defense. Those are different questions from “does AI pay off,” and they are the ones actually in play tonight.

Here is what to watch, line by line, and how to read each outcome.

Alphabet: the capex number is the story, not the EPS

Consensus sits around $2.86–$2.95 EPS on revenue of roughly $116.8–120.2 billion — double-digit growth on both lines, with analysts having raised price targets into the print. A beat on those figures is close to priced in and will not, by itself, move the argument.

The number that matters is capital expenditure. Alphabet’s Q1 capex spiked 107% year over year to $35.7 billion, overwhelmingly directed at AI infrastructure, and full-year 2026 guidance stands at roughly $175–185 billion. Three things to watch there:

Does the guide move? Raising it signals demand visibility strong enough to justify accelerating. Holding it is neutral. Cutting it would be the single most consequential data point of the week — not because less spending is bad, but because it would be the first hyperscaler to blink, and the market would immediately re-rate every supplier priced off that spending.

Does Cloud justify it? Google Cloud grew 63% year over year last quarter with backlog nearly doubling sequentially to over $460 billion. That backlog is the closest thing to a receipt for the capex — it is contracted future revenue against which the infrastructure is being built. If backlog growth stalls while capex climbs, the gap between spending and demand becomes visible for the first time.

Watch depreciation, not just capex. This is the underrated line. Capex hits the cash flow statement immediately; depreciation hits earnings over the following years, typically on a six-year schedule for compute assets. The spending of 2025 and early 2026 is only now beginning to land on income statements. A quarter where revenue growth outpaces the depreciation ramp is a genuinely good sign; one where margins compress as depreciation catches up is the early warning the bears have been waiting for.

Tesla: strong deliveries, and a deliberate cash bonfire

Tesla already reported deliveries: 480,126 vehicles, up 25% year over year against analyst estimates near 406,000 — its strongest growth since Q3 2023. The delivery beat is known. It is not what the print is about.

The number to watch is free cash flow, which the street expects to swing to roughly negative $3.25 billion, against nearly positive $5.6 billion a year ago, as quarterly capex jumps to about $6.7 billion. Full-year 2026 capex guidance is approximately $25 billion, up from $8.5 billion in 2025 — roughly a tripling.

That money is going three places at once: Optimus humanoid production, AI data center buildout, and the Cybercab ramp. Which means Tesla’s quarter is a cleaner test of a specific proposition than Alphabet’s. Alphabet is spending on infrastructure to serve demand that already exists and is partly contracted. Tesla is spending on infrastructure to serve products that do not yet meaningfully sell — a robot without a market and a robotaxi without regulatory clearance at scale. Both may work. They are not the same risk.

The question on the call is therefore narrative, not arithmetic: does management give a date, a unit target, or a revenue figure for Optimus and Cybercab that is specific enough to model? Vague reassurance against a $3 billion quarterly cash burn is a materially weaker answer this quarter than it was last quarter, because the burn is now real rather than projected.

Intel: the one where a single percentage point decides it

Intel is expected to post roughly $0.22 adjusted EPS on $14.42 billion in revenue, against its own guidance of $13.8–14.8 billion and about $0.20. The stock is up roughly 163% year to date and options are pricing a swing on the order of 12–15%, implying tens of billions in market value hanging on the print.

Two lines matter more than the headline.

Gross margin, with 39.5% as the threshold. Above that level suggests 18A yields, pricing and factory productivity are improving faster than expected — the thesis working. Below it suggests the ramp is expensive and the turnaround is being financed by patience rather than economics.

Foundry losses, and whether they are narrowing. Last quarter Intel Foundry reported $5.4 billion in revenue but only $174 million of that came from external customers, on an operating loss around $2.4 billion. That external number is the honest measure of whether anyone actually wants to buy Intel’s manufacturing. The supporting evidence is real — 18A yields reportedly improved to roughly 85% from 65% the prior quarter, and Apple and Microsoft have been named as initial 18A design partners — but design wins are not revenue. Watch whether external foundry revenue starts compounding off that tiny base, and whether the segment loss shrinks as 18A volume rises.

Reading the three together

The useful frame is that these three companies occupy different positions on the same spectrum: how far is the spending from the revenue?

Alphabet is spending against contracted backlog — the shortest distance. Intel is spending against manufacturing capacity with named design partners but negligible external revenue — a longer distance, with an execution milestone (yields) that is at least measurable. Tesla is spending against products that do not yet exist in the market — the longest distance, and the one where narrative substitutes for data.

If all three deliver, the AI capex thesis survives another quarter and the semiconductor complex — including the Korean memory makers selling into exactly this spending — gets a reprieve after a bruising month. If Alphabet trims guidance or Tesla’s burn arrives without a credible timeline attached, the market has a concrete reason to question the spending rather than an abstract one, and that is a meaningfully different market than the one we have had.

The likeliest outcome, for what it’s worth, is messier than either: solid results with at least one uncomfortable line item, and an argument that continues. The capex cycle is too large and too committed to be resolved by three prints. But it can start being repriced by them, and that is what the next 48 hours are actually about.

What to check when the numbers land

Five specifics, in order of how much they’d move the argument:

Alphabet’s FY26 capex guide — raised, held, or cut (cut is the market-moving case)
Google Cloud backlog — still compounding off $460B, or flattening
Tesla free cash flow — how close to the expected −$3.25B, and whether Optimus/Cybercab get dated targets
Intel gross margin vs 39.5% — the single cleanest read on whether 18A economics work
Intel external foundry revenue — any meaningful move off $174M means the foundry story is becoming real

This article is for informational purposes and is not financial or investment advice. All figures are consensus estimates, company guidance and prior-quarter reported results as of publication, ahead of the actual releases; they will be superseded by the reported numbers. Verify against primary sources before acting on any of it.

Sources

This article may contain affiliate links. Nothing here is financial or investment advice.