Meta's spending spree just cost it a quarter of its value

Meta Platforms' shares fell by twenty-four per cent on 27 October, wiping out eighty-five billion dollars of market value, after management said capital spending in 2023 would rise to between $34 billion and $39 billion. Earnings reports have a cold clarity when the market reprices a company's cash flows overnight. Meta's cash generation is being eaten away by weaker advertising and by privacy restrictions on its platforms. For a company in that position, committing forty billion dollars to speculative virtual-reality infrastructure looks less like vision than like a balance sheet losing touch with economic reality.

For the past decade Silicon Valley's biggest platforms enjoyed an unusual luxury. Their core software businesses earned returns on invested capital that were close to infinite. Distributing code to another hundred million users needed almost no physical investment, so extra revenue fell straight through to free cash flow. That cash paid for large buybacks and generous bets on moonshots without worrying institutional shareholders. Those days are over.

A year of rebranding and repricing

The company changed its name from Facebook to Meta a year ago this week, signalling that the virtual worlds of the metaverse would be its next platform. Investors were patient at first. That patience broke in February, when the company reported its first-ever fall in daily users and warned that Apple's privacy changes would cost it about $10bn of revenue this year; the shares dropped by a quarter in a day, then the largest one-day loss of market value for any American company.

Since then the news has only worsened. Revenue fell year on year in the second quarter for the first time since the company went public, and fell again in the three months to September. The division that houses the metaverse effort, Reality Labs, lost more than $10bn last year and is losing money faster this year, and management has said those losses will grow significantly in 2023. Days before the results, Altimeter Capital, a shareholder, published an open letter urging the company to cut headcount by at least a fifth and cap metaverse spending at $5bn a year. The capital-spending guidance was, in effect, the reply.

Servers cost real money

Meta's predicament reflects a move from the variable costs of software to the fixed costs of heavy infrastructure. Building giant artificial-intelligence computing clusters and its own virtual-reality hardware requires large physical assets, from advanced chips to dedicated data centres and big electricity contracts. Engineers can be laid off when advertising revenue cools. Chip orders and data-centre construction contracts cannot be cancelled without heavy penalties.

Meanwhile the business that pays for all this, the advertising network across Facebook and Instagram, is under unusual pressure. Apple's privacy changes have permanently weakened the targeting of digital display ads, and rival short-form video platforms are winning users' attention while earning less revenue per user. Guiding to a big rise in capital spending just as revenue starts to shrink invites investors to value the company on industrial multiples, not platform ones.

The squeeze shows up in the arithmetic of free cash flow. Operating cash flow depends on advertising prices that are falling in a slowing economy, while capital spending is set a year or more in advance. When the second number rises and the first flattens, the residual that funds buybacks shrinks from both ends at once, and it is that residual, not revenue, on which the shares have been priced.

Higher rates, lower multiples

Share prices are, in the end, discounted cash flows. When money cost nothing, investors were happy to give a positive terminal value to distant ventures that had yet to prove they could make money. With the ten-year Treasury yield comfortably above 4 per cent, every dollar spent on assets that produce nothing is discounted heavily.

The effect is sharpest for projects whose payoff lies furthest away. A consumer metaverse that may earn its return a decade out is exactly the kind of long-duration asset that loses most when the discount rate rises, in the same way that a thirty-year bond loses more than a two-year note. Meta is, in effect, lengthening the duration of its cash flows just as the market is punishing duration everywhere.

Owners without a vote

Meta's refusal to adjust its spending to a worsening economy is a serious governance problem. Its dual-class share structure means outside shareholders have no voting power to force a change of strategy or demand that cash be returned. Mark Zuckerberg's class B shares give him a majority of the votes, so the board answers, in practice, to him. Fund managers who dislike the plan have only one way to say so. They can sell the shares and put the money into companies that treat cash with more respect.

There is a precedent for what happens when an industry builds capacity ahead of demand on borrowed conviction. At the turn of the century telecoms companies laid fibre-optic cable on the assumption that internet traffic would justify it quickly; demand did arrive, but years later and for other companies' benefit, after the builders had written off much of the investment. Meta has no debt problem of the kind that sank those carriers, but the lesson about timing applies.

The case for the spending

The strongest defence of the plan is that much of the extra capital goes into computing for artificial intelligence, not headsets. Better machine learning is how Meta hopes to rebuild ad targeting without the tracking data Apple has taken away, and how it ranks the short-video Reels it needs to match TikTok. If that works, part of the spending protects the core business instead of betting on a new one.

That argument has merit, and it is why the shares are not priced for collapse. The difficulty is that management has not separated the two bets in a way shareholders can audit, so the market assigns the whole sum the discount it would apply to the most speculative part. Clearer disclosure of how much goes to advertising infrastructure, and how much to Reality Labs, would be the cheapest way to narrow that discount.

Meta's fall marks the end of growth at any cost for the biggest technology companies. If Meta spends between $34 billion and $39 billion next year with no clear payback, investors will keep valuing it more like a capital-hungry utility than a software monopoly, and the discount will stay until the spending falls. Silicon Valley has found that the rules of corporate finance apply to founders with grand visions, too, and the fourth-quarter results due early next year will show whether Meta cuts costs before its shareholders cut it further.