{"id":845,"date":"2026-10-31T16:24:17","date_gmt":"2026-10-31T16:24:17","guid":{"rendered":"https:\/\/orioncapital.lt\/?p=845"},"modified":"2026-10-31T16:24:20","modified_gmt":"2026-10-31T16:24:20","slug":"ai-cash-burn-1-trillion-week-october-26-30-2026","status":"publish","type":"post","link":"https:\/\/orioncapital.lt\/lt\/ai-cash-burn-1-trillion-week-october-26-30-2026\/","title":{"rendered":"The $1 Trillion Burn"},"content":{"rendered":"\n

Three of Four AI Giants Now Spend More Than They Earn<\/strong><\/p>\n\n\n\n

Alphabet, Amazon and Meta are all burning cash. The Federal Reserve is leaning toward its second rate rise of the year. Oil crossed $117. All of it landed in five sessions.<\/em><\/p>\n\n\n\n

October 26\u201330, 2026 | Ieva Bru\u017eait\u0117, Orion Capital<\/p>\n\n\n\n

What happened<\/strong><\/h1>\n\n\n\n

A trillion dollars did not burn \u2014 share prices were repriced. The $9.07 billion that left Alphabet\u2019s bank account did.<\/p>\n\n\n\n

Friday, 30 October was the worst session for US equities since June, and more than $1 trillion of technology market value burned off between the opening bell and the close. The Nasdaq Composite fell 2.86%, the Philadelphia Semiconductor Index fell 7.2%, and Intel lost 28.6% in a single day \u2014 its steepest fall since 1974. [1][2][3]<\/p>\n\n\n\n

The weekly numbers tell a quieter story, and the gap between the two is the point. Over the full week the Nasdaq lost 2.0% and the S&P 500 lost just 0.6%, while the Russell 2000 actually gained 1.3%. Friday was a concentrated repricing of one trade, not a broad retreat from equities. [1][4]<\/p>\n\n\n\n

Two separate forces arrived in the same week and pushed on the same valuations from opposite ends. The first was a cash burn inside the AI buildout that three of the four largest cloud builders can no longer present as a timing difference: Alphabet, Amazon and Meta all reported negative free cash flow. The second was a war premium in energy that is now feeding a central bank which has already stopped cutting and started raising. One compresses the numerator of a valuation. The other raises the denominator.<\/p>\n\n\n\n

For two years the market financed AI on description.<\/em><\/p>\n\n\n\n

This week it began asking for bank statements.<\/em><\/p>\n\n\n\n

The week in numbers<\/strong><\/h1>\n\n\n\n
Index<\/strong><\/td>Close<\/strong><\/td>Friday<\/strong><\/td>Week<\/strong><\/td>Note<\/strong><\/td><\/tr><\/thead>
Nasdaq Composite<\/td>25,437.00<\/td>\u22122.86%<\/strong><\/td>\u22122.0%<\/td>\u22125.3% in October<\/td><\/tr>
S&P 500<\/td>7,580.07<\/td>\u22121.82%<\/td>\u22120.6%<\/td>Third weekly decline<\/td><\/tr>
Dow Jones<\/td>51,770.31<\/td>\u22121.09%<\/td>\u22120.9%<\/td>Cushioned by defensives<\/td><\/tr>
Russell 2000<\/td>2,811.84<\/td>\u22121.66%<\/td>+1.3%<\/strong><\/td>Only index up on the week<\/td><\/tr>
PHLX Semiconductor<\/td>\u2014<\/td>\u22127.2%<\/strong><\/td>\u2014<\/td>The real epicentre<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
Elsewhere<\/strong><\/td>Level, 30 Oct<\/strong><\/td>Friday<\/strong><\/td>Note<\/strong><\/td><\/tr><\/thead>
10-year Treasury<\/td>4.93%<\/td>\u22126 bp<\/td>Near 5% with a hike priced for December<\/td><\/tr>
Brent crude (Dec)<\/td>$117.64<\/td>+2.4%<\/td>+13.6% in October; fourth monthly gain<\/td><\/tr>
WTI crude<\/td>$104.12<\/td>+2.08%<\/td>+15.2% in October<\/td><\/tr>
Gold<\/td>~$3,885<\/td>\u22123.6%<\/td>One-year low; silver \u22126%<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

Index levels and daily changes are closing figures for 30 October; weekly and monthly figures are rounded. [1][3][4][5]<\/em><\/p>\n\n\n\n

$1 Trillion that moved the market<\/strong><\/h2>\n\n\n\n

Three of the four largest cloud builders reported negative free cash flow in the same week. That has not happened before in this cycle, and it is the reason a quarter of broadly strong operating results produced a selloff. The burn is no longer a forecast in a bear note. It is in the filings.<\/p>\n\n\n\n

Company<\/strong><\/td>Quarterly capex<\/strong><\/td>Free cash flow<\/strong><\/td>Year earlier<\/strong><\/td>2026 capex<\/strong><\/td><\/tr><\/thead>
Alphabet<\/td>$51.4bn<\/td>\u2212$9.07bn<\/strong><\/td>+$24.5bn<\/td>$195\u2013205bn<\/td><\/tr>
Amazon<\/td>\u2014<\/td>\u2212$42.4bn<\/strong><\/td>+$14.8bn<\/td>~$220bn<\/td><\/tr>
Meta<\/td>\u2014<\/td>\u2212$2.12bn<\/strong><\/td>\u2014<\/td>$135\u2013145bn<\/td><\/tr>
Microsoft<\/td>$45.2bn<\/td>+$18.6bn<\/strong><\/td>+$25.7bn<\/td>>$50bn next Q<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

Alphabet’s and Microsoft’s figures are for the September quarter. Amazon’s is a trailing-twelve-month figure, not a quarterly one \u2014 the distinction matters and is often lost in summaries. Meta’s is for its September quarter, reported 21 October. [6][7][8]<\/em><\/p>\n\n\n\n

Read the operating lines on their own and the quarter looks excellent. Google Cloud revenue grew 103% to $30.71 billion. AWS grew 39% to $46.0 billion, its fastest rate in nearly five years, and AWS operating income rose from $11.4 billion to $17.8 billion. Azure grew 48%, ahead of the 44% expected, and Microsoft’s commercial backlog reached $677 billion, up 73%. These are not the numbers of an industry running out of demand. [6][7]<\/p>\n\n\n\n

The market sold them anyway, because the cash statement said something the income statement did not. Alphabet burned $9.07 billion in the September quarter against positive $24.5 billion a year earlier, spent $51.4 billion on capital against $24.0 billion, and covered the gap by issuing $28.7 billion of senior unsecured debt and raising a further $12.4 billion through at-the-market equity sales. A company with Alphabet’s balance sheet does not do that for want of money. It does it because the buildout now consumes cash faster than one of the most profitable businesses in history generates it. [6]<\/p>\n\n\n\n

Amazon’s burn is larger and longer-running: negative $42.4 billion on a trailing-twelve-month basis, against positive $14.8 billion a year before. Meta turned negative in its September quarter at $2.12 billion. Three companies, three different stages of the same problem. [6][7]<\/p>\n\n\n\n

Spending more than you earn is not automatically a mistake.<\/em><\/p>\n\n\n\n

It is a claim about the future that now has to be defended every ninety days.<\/em><\/p>\n\n\n\n

Alphabet’s headline earnings miss \u2014 $2.12 per share \u2014 was mostly an $8.7 billion loss on the revaluation of its equity investments, including its SpaceX stake. That is a mark-to-market item, not an operating failure, and I would not weigh it heavily. Search revenue grew 12% to $63.4 billion. The cash flow line is the one that deserves the attention. [6]<\/p>\n\n\n\n

Microsoft is the control case, and it is the most useful company in the group this week. It is spending on the same scale \u2014 $45.2 billion in the quarter and more than $50 billion guided for the next \u2014 and it is still not burning cash, producing $18.6 billion of positive free cash flow with a backlog up 73%. The difference is not that Microsoft is spending less. It is that Microsoft has pre-sold more of what it is building. [7]<\/p>\n\n\n\n

Intel: a different kind of risk, wearing the same label<\/strong><\/h2>\n\n\n\n

Intel fell 28.6% on Friday to $65.76, the worst single session in the company’s listed history since 1974. The cause was guidance: fourth-quarter revenue of $15\u201316 billion against roughly $17.1 billion expected, and adjusted earnings of $0.25 against $0.43, alongside a warning on cash flow. The third-quarter results themselves had beaten. [2][5]<\/p>\n\n\n\n

Intel is not an AI capital-expenditure story. It is a company with a weak PC market, an expensive foundry ambition and a government shareholding that lost roughly $12 billion of value in a day. It fell on the same Friday as the hyperscalers and it dragged AMD down 10.9% and Nvidia down 4.2% with it. [3][5]<\/p>\n\n\n\n

That is the behaviour worth noting. When a semiconductor index falls 7.2% in a session, it is not discriminating between a memory supplier with sold-out capacity, a networking business with a multi-year order book, and a manufacturer missing its own guidance. The label is doing the work that analysis should be doing. Periods when correlations go to one are generally where mispricing is created, in both directions.<\/p>\n\n\n\n

The Fed is not pausing. It is tightening.<\/strong><\/h2>\n\n\n\n

This is the part of the week most likely to be misread, and it changes the arithmetic underneath every long-duration asset.<\/p>\n\n\n\n

On 28 October the Federal Open Market Committee held the target range at 3.75\u20134.00%. The vote was 10\u20132. Both dissenters, Governor Lisa Cook and Cleveland Fed President Beth Hammack, wanted a quarter-point increase. The hold followed a quarter-point rise at the September meeting \u2014 the Committee’s first increase since 2023, delivered under Chair Kevin Warsh with a stated commitment to a timelier return to 2% inflation. On Friday, St. Louis Fed President Alberto Musalem and Richmond’s Tom Barkin both said higher rates may be needed. [4][9][10]<\/p>\n\n\n\n

Futures now price roughly an 83% probability of a rise at the 8\u20139 December meeting, up from 73% on 26 October and 42% a month before that. The European Central Bank is priced for a December increase as well. [3][9]<\/p>\n\n\n\n

A hold inside a cutting cycle and a hold inside a hiking cycle<\/em><\/p>\n\n\n\n

are the same headline and opposite facts.<\/em><\/p>\n\n\n\n

The data give the Committee room to move. Third-quarter real GDP grew at a 3.4% annualised rate, up from 2.2%, and real final sales to private domestic purchasers \u2014 the cleanest read on underlying demand \u2014 grew at 4.9%. On inflation the picture is genuinely mixed, and I would not present it as settled: headline PCE rose 0.5% in September and 3.6% over the year, with core at 3.1%, while the third-quarter annualised rates cooled to 2.1% headline and 2.5% core. The quarterly run-rate looks close to target. The monthly print does not, and it was taken before oil moved 15% in October. [4][11]<\/p>\n\n\n\n

For an investor the mechanism matters more than the forecast. A ten-year Treasury at 4.93% is the rate against which a data centre earning its return in 2030 is discounted. Alphabet funding capital expenditure with $28.7 billion of new debt is financing a burn into a rising curve. When the discount rate goes up at the same time as the cash flows being discounted turn negative, multiple compression does not require anyone to doubt the technology.<\/p>\n\n\n\n

The war premium nobody has priced properly<\/strong><\/h2>\n\n\n\n

Brent settled at $117.64 on Friday, up 13.6% in October; WTI settled at $104.12, up 15.2%. That is the fourth consecutive monthly gain, and it is not being driven by demand. [5][12]<\/p>\n\n\n\n

Four tankers were reported on fire in and around the Strait of Hormuz, which Iran’s Revolutionary Guard attributed to mines on an unauthorised route; UK Maritime Trade Operations reported explosive devices attached by a small boat; US Central Command disputed the mine claims and said the main transit lanes were clear. Separately, Houthi forces claimed strikes on Saudi energy infrastructure including the 550,000 barrel-per-day Ras Tanura refinery. The United States has ordered non-emergency staff out of Saudi Arabia. [12][13]<\/p>\n\n\n\n

The shipping market is where the stress is visible without ambiguity. Very large crude carrier rates on the Gulf-to-China route reached $1.33 million a day, an all-time high. Freight has risen from roughly 3% of the delivered cost of a barrel to roughly 27%. War-risk premiums run at 6\u201310% of hull value per transit, and no protection-and-indemnity club has restored Gulf war-risk cover since the major clubs withdrew it in March. Commercial transits through the strait have fallen to single digits a day against a pre-war baseline near 138. [13]<\/p>\n\n\n\n

Traders are positioning for the possibility of US strikes after the 3 November midterms. Saxo Bank expects crude to settle at $70\u201390 once the conflict ends, against $60\u201370 before it began \u2014 which is to say that even the resolution case carries a permanently higher floor. [12]<\/p>\n\n\n\n

I draw a narrow conclusion from this rather than a dramatic one. Energy was the strongest S&P 500 sector of the year, Exxon and Chevron touched intraday records on Friday on earnings beats, and the market has so far treated the conflict as an energy-sector event. It is also an inflation event, which makes it a rates event, which makes it an AI-valuation event. That chain was the connective tissue of this week and very little commentary drew it. [3][5]<\/p>\n\n\n\n

Where the money actually went<\/strong><\/h2>\n\n\n\n

The Russell 2000 rose 1.3% for the week while the Nasdaq fell 2.0%. Energy gained. Defensive names \u2014 Mondelez, Colgate, Walmart, Coca-Cola, Procter & Gamble \u2014 were firm. Financials lagged. Gold fell 3.6% to a one-year low and silver fell 6%. [1][3][5]<\/p>\n\n\n\n

The gold move is the one I would flag, because it cuts against the easy story. If this week had been a broad flight from risk, precious metals would not have had their worst session in months on the same day. Money did not leave the market. It moved from assets whose value sits far in the future to assets that produce cash now. That is a duration rotation, not a panic.<\/p>\n\n\n\n

One week does not establish a rotation. Three would. It is worth watching rather than acting on.<\/p>\n\n\n\n

Orion Capital lens | Chips \u2192 Compute \u2192 Power<\/strong><\/h2>\n\n\n\n

Our infrastructure framework is unchanged by this week, but the week sharpened which question inside it matters. AI expansion needs semiconductors and servers, and it equally needs grid connections, transformers, electrical equipment, cooling, construction labour and financing. A delay in any one link moves revenue recognition and cash flow along the whole chain.<\/p>\n\n\n\n

Three questions now matter more than headline capital expenditure:<\/p>\n\n\n\n