Meta Answered the Funding Question with Equity. The Market Sold the Answer.

Thursday, October 22, 2026, written at 9:15 a.m. Eastern, fifteen minutes before the open. Every future, yield, currency and commodity level in this note carries that stamp, and the European and Asian figures are intraday or local closes rather than settled American marks. Weekly jobless claims landed at 8:30 and came in at 194,000 against 200,000 expected. Tomorrow brings September PCE and the start of the Fed blackout, so today is the last full session before the committee goes quiet into October 28.

Top Call: Meta Chose Equity Over Debt, and That Is the Concession

Meta reported $64.6bn of revenue last night, up 26% and ahead of a $63.34bn consensus, with Q4 guided to $73bn to $76bn. It also raised the floor of its 2026 capital spending range to $135bn from $130bn, spent $34.91bn in the quarter alone, and posted free cash flow of negative $2.12bn. Then, in the same release, the board approved a new non-voting Class C share class, distributed as a stock dividend effective January 19, with filings stating the shares may be used to raise capital.

That last item is the whole note. For two weeks this desk has argued that the binding constraint in this cycle is the cost of capital rather than the level of demand. Meta did not argue with that. It built a financing instrument around it. When your incremental debt prices off a spread that Goldman’s traders have marked at the wides of the year, and your free cash flow has gone negative, you fund the plan with stock instead. Zuckerberg keeps control through the voting structure, and the dilution lands on everybody else.

Our read is that this is the most honest disclosure of the week and the market treated it correctly within hours. Nasdaq 100 futures are down 1.85%, the VIX is up 11% to 18.54, and Meta itself gave up an early gain to trade down about 1.5%. Equity funding tells holders exactly what it costs them. Debt funding does not, which is why it took a rating agency to point out on Monday that nobody can see all of it.

The Sequence Is the Argument

Compress the last five sessions. Monday, S&P Global said the largest technology companies almost certainly carry more artificial intelligence debt than the agency can measure. Tuesday night, SpaceX guided fourth quarter capital spending to $35bn to $40bn against $19.6bn just spent. Wednesday night, Meta created a share class whose stated purpose includes raising capital.

Each of those is a funding event dressed as something else. What moved in under a week is not the demand case, which is holding up on every revenue line anybody has published. It is the location of the question. On Monday the funding problem belonged to the ratings agencies and their estimates. By last night it was inside a capital structure, approved by a controlling holder, with a date attached.

One caveat we want to state plainly, because it cuts against the strength of our own framing. Meta has not sold a share. The Class C dividend is functionally a two for one split that creates a currency for later issuance, and the company has published no plan or timeline to use it. So what exists today is optionality, not dilution. We still think it matters, because building the vehicle is the expensive admission and using it is the cheap part, but anyone reading this as a completed equity raise is ahead of the facts.

Tesla Softened the Language, and the Softening Is a Deferral

Our pre-committed test for last night was whether at least one of Meta, Tesla or IBM would soften capital expenditure language. Tesla did. Management cut 2026 capital spending guidance to roughly $20bn from more than $25bn, with the CFO citing efficient capital deployment. On its stated terms the test resolves yes, and we mark it that way.

The resolution is less useful than the test assumed, and we should say so rather than bank the point. Tesla’s third quarter capital spending more than doubled to $3.73bn. Free cash flow fell 62% to $1.53bn and net income fell 29% to $787m as artificial intelligence and Optimus costs grew faster than gross profit. On the call Musk said spending runs lower than expected only for now, grows meaningfully in 2027, and reaches tens of billions in 2028, with Terafab equipment alone possibly passing $100bn over time. Alongside that, the company arranged capacity to borrow up to $30bn.

A company that lowers a twelve month guide, doubles quarterly spending, promises tens of billions in two years and lines up $30bn of borrowing capacity has not changed its ambition. It has changed the calendar. Revenue was roughly flat near $28.05bn, adjusted earnings of $0.48 beat $0.43, services grew 33% to $4.61bn and full self driving subscriptions rose 67%. The shares are down about 5%. The honest conclusion from the test we wrote is the one we pre-committed to: our cost of capital frame has changed funding structures and timing, not the size of the plans.

The Tape: This Is Positioning, Not a Macro Signal

Look at what went down together. Nasdaq futures are off 1.85%, gold is down 0.8% to $4,087, silver is down about 3.4%, copper is down 2% to a one month low, and bitcoin is down to roughly $81,983. Treasuries rallied, with the ten year five basis points lower at 5.07% and the two year at 4.74%. Energy is the only sector higher.

Equities, precious metals and industrial metals falling on the same morning that bonds rally is usually not a growth scare. It is books getting smaller. Goldman’s risk appetite indicator went negative on Tuesday and hedge funds set a record for net selling of global technology across two weeks, so the marginal seller here has been reducing for a fortnight and the Meta filing gave the rest of the market a reason to join.

Asia carried the move. The Nikkei closed down 2.65% at 62,908, Hong Kong fell 3.1% to 22,919.11, the Hang Seng Tech index lost 5% to a two year low, and the Kospi dropped more than 4% with SK Hynix off 5.7% and Samsung down about 4%. The CSI 300 fell 1.84% and is now roughly 20% below its June peak, a tenth consecutive weekly decline and a bear market on the conventional definition.

We would separate China from the rest of that list. Korea and Japan fell because they sell the hardware whose buyers just published a funding problem. China fell for its own reasons, including a long consumption slump, fading stimulus expectations and a tax deadline that pushed large holders to sell. Reading the CSI 300 as a verdict on artificial intelligence spending is the easy mistake today.

Earnings: Beats That Did Not Hold

IBM is the sharpest illustration of the regime. Software grew 8% to $7.8bn and beat, adjusted earnings beat, the free cash flow outlook went up and the software growth floor was raised toward 7% to 8%. The shares traded up as much as 7.4% to $269.50 and then went out near $238. What killed it was the full year revenue growth guide staying at 4% to 5% in constant currency. Holders wanted the beat flowed through to the year and did not get it, and UBS noting that less than half the delayed deals have been recaptured did not help. Beating the quarter has stopped being sufficient for three weeks running now.

Lam Research produced record revenue of $8.22bn and fell about 6%. The December guide of $8.2bn to $9.0bn brackets the $8.53bn consensus, with adjusted earnings of $2.18 to $2.48 against $2.31. That is a fine guide in isolation. In a tape where buyers are asking who pays for the equipment, a midpoint beat with a low end below consensus gets sold. We would note the contrast with Texas Instruments, which guided fourth quarter revenue to $5.8bn to $6.3bn against $5.73bn expected on data centre demand. Analog, with short cycle orders and no financing story attached, is where money is comfortable this morning.

Barclays deserves attention for a reason unrelated to its quarter, which was mediocre. Pre-tax profit rose 12% to £2.3bn in line with forecasts, group income of £7.4bn missed, return on tangible equity dropped to 11.7% from 16.1%, and equities trading rose 47% while fixed income fell 19%. The item that matters is the announcement of roughly 5,000 job cuts, about 5% of the workforce, with 1,000 already gone and roughly 60% of the reductions in support functions including developers. This is the first large, specific, dated claim by a major employer that artificial intelligence is removing headcount at scale. Ken Griffin’s complaint on Tuesday was that corporations cannot identify clear productivity gains from their invoices. Barclays just put a number in a public filing. One bank is not a trend, and cost programmes have worn fashionable labels before, but if the productivity case is ever going to show up it shows up here, in employers’ headcount, not in vendors’ revenue.

Energy: Four Tankers and a Five Month High

Brent trades at $112.27, up 3.2% and the highest since May. WTI is near $99.80. The Revolutionary Guard said it attacked four vessels near the Strait of Hormuz using a route Tehran rejects, and said three other ships on the same route struck sea mines and had to stop. Renewed Houthi attacks on Saudi targets added to it. Trump has denied that Washington is negotiating with Tehran, and there is no published timetable for reopening the strait.

Kpler still puts Hormuz flows near 82% of pre-war levels, which is the number to hold onto. Physical barrels are mostly moving. What is repricing is the probability that they stop. That keeps our split intact: a 2027 crude surplus sitting underneath a present distillate and freight shortage, with the risk premium now attached to transit rather than production. The one change worth flagging is that crude is climbing on a day when gold is falling, which is unusual enough to suggest the oil move is a supply story in its own right rather than part of a general hedge against disorder.

Europe: France Now Pays More Than the United States

The French ten year rose to 5.09%, briefly above 5.11%, the highest since 2002, and is now trading above the United States ten year for the first time since 2013. The spread over Germany reached about 169 to 170 basis points, the widest since 2011 and on some measures the widest of the single currency era. Equities finally joined the bonds, with the CAC 40 down about 1.2% to 1.4%, the DAX off 1%, the FTSE 100 down 0.7% and the Stoxx 600 down 0.95% at 638.38, with European banks off 2.7%.

We wrote yesterday that bond and equity investors in France could stay in disagreement for a while and that the thing which would force convergence is a bank funding cost. European banks are the weakest sector in Europe this morning. That is the mechanism arriving on schedule, and it is the reason tomorrow’s Moody’s review carries more weight than a ratings decision normally would. A sovereign whose ten year yields more than a United States Treasury while running a 5.4% deficit, a paralysed budget process and a national day of action booked for November 5 is being priced as a credit, not a rate.

Elsewhere, Hermès fell about 7% on results, Nestlé added 1.2%, LSEG rose almost 4% on a record quarter with raised guidance, and Nokia beat with artificial intelligence related sales up 123% and still fell 5.6%, which is the European version of this morning’s American pattern.

Also On The Tape

• DeepSeek has raised about $15bn, near RMB 100bn, after lifting its target twice, with Tencent and CATL among the largest investors and an accelerated path toward a Shanghai STAR Market listing. The company whose efficiency claim knocked the complex on Monday has now raised capital on the strength of it
• Western Digital reports after the close tonight into a market where nearline capacity is sold out through calendar 2026, Seagate calls its nearline supply fully allocated, and Morgan Stanley models a 400 exabyte industry shortfall across 2027 and 2028 against 300 exabytes this year
• EQT agreed to acquire Qiagen for about EUR 9.6bn, or $10.8bn, at EUR 46.50 a share, a premium near 8.6%
• CBRE agreed to buy Onnec for up to $1.01bn to expand its data centre infrastructure services
• Samsung Electro-Mechanics is reported to be receiving about $2.8bn from Tesla and Nvidia for artificial intelligence chip substrates
• Jobless claims fell to 194,000 against 200,000 expected, a three month low, which sits awkwardly beside September payrolls of 29,000 and tells you employers are slowing hiring without shedding staff
• Fed pricing for October 28 holds near 86% to 89% for no change, while a December hike is now close to a consensus view, with one survey reporting all 100 economists polled expecting one
• Pre-market chip moves: AMD down 4.5%, Broadcom down 2.9%, Oracle down 2.8%, Nvidia down 2.4%, Micron down 3.4%, Super Micro down 4.5%, Corning down 4.5%

The Record

Three marks and one new test.

The capital expenditure test we pre-committed on Monday and restated yesterday resolves yes on its stated terms, because Tesla cut its 2026 guide. The addendum we added yesterday also fired, because Meta raised rather than softened. We said in advance that if Meta raised, the honest conclusion would be that our cost of capital frame had changed funding spreads rather than behaviour. We are keeping that, with one amendment that makes it less flattering to us: Meta did not merely keep spending through a higher cost of capital, it went and built a cheaper funding channel. That is behaviour changing, just not in the direction we were looking.

We were also wrong about timing in a way worth owning. Yesterday we wrote that the market was pricing SpaceX revenue and deferring the financing question to the next funding round. The deferral lasted about twenty four hours, and it took a share class filing rather than a funding round to end it. The direction was right. The duration was not, and we would have carried risk for a day on that assumption.

The Volta secondary test reaches its final day tomorrow. We said the loan should trade at or below its 95 cent issue price through its first week of secondary trading, with a print above 96 meaning we overstated how expensive this funding has become. There are still no reliable quotes anywhere we can source. Absent one tomorrow we mark it unresolved rather than claim it, as stated.

New test for tonight. Western Digital is the cleanest remaining read on whether the physical bottleneck argument we have leaned on since October 13 is still intact. If it is, the company should guide fiscal second quarter revenue above consensus and reiterate that capacity is committed beyond calendar 2026. If it guides in line, or signals any normalisation in nearline pricing, then the supply shortage leg of our argument is weaker than we have been saying and we will say so tomorrow rather than quietly fold it into a sentence about demand.

Key Events Today

• 8:30 a.m. Weekly jobless claims, released at 194,000 against 200,000 expected
• Pre-market Procter & Gamble, Honeywell, Comcast, Union Pacific, Blackstone, Lockheed Martin, Valero, Norfolk Southern, Dow, Nasdaq Inc, American Airlines and Spotify
• After the close Western Digital, Newmont, Deckers, VeriSign, Bloom Energy and Hartford
• Friday September PCE, the last inflation print before the Fed, and the start of the blackout
• Friday Moody’s review of France, with the ten year above the United States for the first time since 2013
• October 27 and 28 FOMC, a hold priced near 89%
• Watching Hormuz transit, any secondary quote on the Volta loan, European bank funding costs, and whether a second large employer follows Barclays in attributing headcount reductions to artificial intelligence