Big Tech AI Capex: Who Can Fund It

TSMC’s October revenue jumped 59.7% to a record NT$586.89 billion, just days before an AI-chip selloff. TSMC reported that monthly print on Nov. 10, 2026, and the sequence was the point: wafer demand from Nvidia (NVDA) and other accelerator buyers was still hot, and the equity tape around the print stopped handing out multiple expansion for every extra dollar of data-center spending. The spending is real, and the multiples have stopped following it. The split that matters is funding capacity.

Company guides are the figures that survive a check. Alphabet (GOOGL) held 2026 spending at $195 billion to $205 billion and the stock still sold off. Amazon (AMZN) is holding roughly $220 billion. Meta Platforms (META) guided $130 billion to $145 billion, up from $72 billion. Microsoft (MSFT) is the cash outlier in that set, with fiscal first-quarter free cash flow of $18.6 billion. All four are building the same thing, and each is paying for it a different way.

Company Spending Guides, Not the Group Headline

Free Cash Flow Is the Split Inside the Spender Group

Microsoft’s fiscal first quarter, ended Sept. 30, 2026, and reported on Oct. 28, produced revenue of $92.3 billion, up 19%, and net income of $36.6 billion. That Oct. 28 release is the clean revenue print. Azure grew 48% to $33.2 billion, which one note described as the first time the company put a public dollar figure on that line, and as the fastest growth in at least two years. Free cash flow was $18.6 billion, $10 billion above estimates in that same note. Property and equipment additions were $39.4 billion, more than double a year earlier, and a separate note said quarterly capital spending landed $6 billion below estimates. Post-print coverage also described a plan for more than $50 billion of quarterly spending ahead. It was a heavy quarter, and cash flow still came in positive by a wide margin.

Alphabet’s cash story in the same week was the opposite sign. A 247 Wall St. comparison put the spending surge at $51 billion and free cash flow at negative $9 billion for a second straight quarter, against Microsoft’s $18.6 billion of positive free cash flow. Use the $18.6 billion official figure; a secondary comparison put Microsoft near $19 billion, close enough that the sign is the same. Another writeup said Alphabet posted its first negative free cash flow since the 2004 IPO, which conflicts with the “second straight quarter” wording. The sign is the shared fact; the streak length is not, so this article uses “negative free cash flow” without a streak count.

The funding mix is what makes the negative print tradable. 247 Wall St. reported that Alphabet raised $41 billion in one quarter, describing stock sales and new debt after capital spending exceeded operating cash flow. The same inventory note said a TPU build had reached $9.99 billion by mid-2026 and was weighing on cash flow. Search advertising still grew: one earnings recap put advertising revenue at $82.0 billion, up 10.6% year over year, slower than 14.4% in the second quarter. Ads cover part of the build, and the rest comes from somewhere else.

Meta’s cash print drew a rating-agency response, not just a multiple reset. Coverage of the quarter put free cash flow at about negative $2.12 billion, and S&P revised the outlook to negative as the 2026 spending bill rose. A later note had the shares dipping toward $600 after the post-earnings tech rout deepened following the midterm results. Wells Fargo cut its price target after the quarter and, in that coverage, modeled $250 billion of capital expenditure in 2027. That $250 billion is an analyst model rather than a company guide. Amazon’s post-earnings commentary treated the $220 billion spending hold as the reason the stock fell on Oct. 29 even with a 39% cloud growth rate. A Nov. 10 Yahoo note said Amazon had by then recouped those post-earnings losses, which shows a path back without settling the cash debate.

If the next set of reports shows Microsoft still the only large spender with clearly positive free cash flow, the group will keep trading as four different funding stories rather than one basket.

TSMC and ASML Still Describe a Shortage, Not a Glut

Supply tightness is the counterweight to the multiple compression. The October foundry number is the cleanest supply fact of the last week. NT$586.89 billion is a record monthly sales print, up 59.7%, and the company tied it to chip demand from Nvidia and other customers. For the September quarter, two writeups agree on revenue of about $46.72 billion and disagree on the growth rate: one said 41.2% year over year, another 50.9%. Use the dollar level. Do not pick a growth rate until a primary filing settles it. A Seeking Alpha note said the quarter beat the top end of revenue guidance by $920 million and described the shares as down 18%, tying a recovery case to the 2nm ramp without giving a yield or a capacity number that can be repeated here. The demand mix supports the same read: a separate account said high-performance computing was more than two-thirds of quarterly revenue, and put the share at 68%.

The stock has not tracked the sales line one-for-one. Another note said the shares were trading around 9% under their record even after another forecast increase, and argued that demand still far above supply was a reason to own Nvidia. That is an opinion tied to a real constraint, with no price target and no date attached. The constraint itself is the part with a receipt: record monthly sales in October, published Nov. 10, into a chip selloff rather than a breakout.

The toolmaker tells the same story. ASML Holding (ASML) is the tool side of the same constraint. The company reported third-quarter 2026 total net sales of 11.7 billion euros and net income of 5.0 billion euros. Coverage of the same quarter put the gross margin at 57.5% and described record results, while one account said earnings beat and revenue missed estimates. A record level and a consensus miss can both be true. The Motley Fool’s Oct. 16 reading said ASML cannot keep up with 2027 or 2028 demand and is lifting manufacturing capacity beyond a prior 30% target, with customers interested in 2029 and 2030 orders.

The foundry and the toolmaker are describing 2027 to 2030 tightness at the same time equity buyers are marking down the customers who have to pay for it. That gap is where the trade sits. The next TSMC monthly sales release is the hard check on whether October was a peak or a floor.

Three Cloud Growth Rates, Three Different Bases

The growth rates that have to carry these budgets are not comparable just because they are all large. Google Cloud revenue hit $30.7 billion, up 103% year over year, in a quarter where the parent missed Wall Street revenue estimates and reported what one headline called its highest-ever quarterly capital spending. Azure’s 48% gain took that business to $33.2 billion. AWS revenue rose 39% to $45.98 billion, which a separate writeup rounded to $46 billion, and that card said Andy Jassy projected growth above 40% in both 2027 and 2028. A 103% rate on a $30.7 billion base, a 48% rate on $33.2 billion, and a 39% rate on $45.98 billion are three different dollar increments. Google Cloud’s percentage lead does not mean it added the most revenue, and it arrived in the quarter the parent had to raise outside money.

Microsoft’s limitation is capacity, not demand: 48% growth and still not enough capacity, with a forward quarterly spending bill described as topping $50 billion. That operating comment carries a financing consequence. If GPU and power delivery slip, the growth rate the market just paid for becomes the miss. Amazon’s limitation is the opposite optics problem. Cloud growth was the fastest in nearly five years on one account, and the stock still fell on Oct. 29 because $220 billion of infrastructure spending was the number investors weighed against it. Meta does not have a public cloud line to point at. Its spending case rests on ads and on products that have not been given a dollar return on this outlay in the sources used here, which is why a negative cash print and a negative S&P outlook landed harder.

Oracle (ORCL) is the cautionary fifth spender, on older but still relevant numbers. A Sept. 21, 2026 account said the stock had plunged 50% even after co-CEOs delivered a $664 billion contract backlog and 121% cloud-infrastructure revenue growth in a year, with first-quarter capital spending of $28.5 billion and negative free cash flow. The backlog is a claim on future revenue, and it did not stop the derating. Anyone treating a multi-year cloud contract as cash should read that sequence before treating Meta’s or Alphabet’s 2027 step-up as self-funding.

The print to watch into the next cloud quarter is whether Azure can stay above 40% while quarterly spending clears $50 billion, and whether AWS can hold the “above 40% in 2027 and 2028” comment once the $220 billion year is in the rear-view mirror.

Apple Opted Out of the Race

Apple is the control case for the intensity argument: a mega-cap of similar scale that chose not to join the race. The point is not that Apple is a foundry customer; it is that the same mega-cap tier contains a funder and a non-funder, and the market is repricing the difference. That choice is useful because the market spent two years treating “tech” and “data-center infrastructure” as the same trade. Seeking Alpha’s news note on Apple (AAPL) said capital spending stayed near its early-2023 baseline through the third quarter of 2026 and under 5% of sales, against Meta above 50% of sales. No dollar spending figure for Apple was stated in that note, so none is invented here. The ratio is enough. Apple shows that a mega-cap can refuse the build and still be a mega-cap. That does nothing to fix a negative cash quarter at a cloud buyer, and Apple is not a foundry customer.

The practical split for anyone running infrastructure budgets, not just a book of stocks, is intensity. Meta above 50% of sales and Apple under 5% are not two points on a smooth curve. They reflect different corporate choices about who owns the data-center bill. Suppliers such as TSMC and ASML get paid either way, as long as the buyers who stayed in the race keep ordering. The equity risk sits with the buyers whose free cash flow has already flipped negative.

The 2027 Meta Test

The clearest test of the funding race is a single 2027 guide. Meta Platforms will guide 2027 capital expenditure above $200 billion by Feb. 28, 2027, because the 2026 range has already moved from $72 billion to $130 billion to $145 billion. The $250 billion 2027 model is not a company guide; it is the trajectory check that makes a sub-$200 billion guide unlikely. That is the same free-cash-flow split in forward form: Meta is already funding the build from a negative cash position, so the 2027 guide is a funding decision, not just a spending number. A guide under $200 billion by that date would falsify the call. The mechanism is the one already sketched: 2026 was the step-up year, and 2027 is another increase, not a pause.

More in-depth coverage from this blog on closely related topics:

Sources and References

Sources cited while researching and writing this article: