Big Tech has a herding problem

The AI Data Centre Rush: Following the Herd

A growing chorus of concern is emerging over the spending habits of Big Tech. An increasing number of the world’s largest companies are committing tens of billions of dollars to artificial intelligence data centres not because they have a clearly mapped path to profit, but because their rivals are doing the same. This phenomenon, often described as herding, risks destroying shareholder value as businesses pivot away from their core strengths into capital‑intensive infrastructure plays without a concrete plan to monetise the investment.

Meta is the poster child for this shift. The social media giant is burning through its cash reserves, with free cash flow expected to turn negative this year after peaking at $52bn in 2024. Founder Mark Zuckerberg argues that “there’s just nowhere near enough compute” and that near‑term capacity is more valuable, but his explanations turn vague when the conversation shifts to exactly how this will generate a return. Meta is developing coding tools for business customers and has floated the idea of renting out computing power, yet Zuckerberg himself admits that selling to enterprises is a “different muscle” for the company.

The trend is not limited to Meta. Tesla, a carmaker, is building its own semiconductor fabrication plant in Austin, while rocket company SpaceX reportedly spent 86% of its capital expenditure budget on AI computing. Elon Musk claims that SpaceX’s rocket‑building expertise will “yield tremendous benefits” when applied to data centres, but industry observers see these moves as expensive detours from core business lines, driven more by fear of missing out than by a distinctive competitive edge.

Why Meta’s Pivot Leaves It Vulnerable

Meta’s Catch‑Up Costs: No Cloud Legacy to Lean On

Unlike its peers, Meta has no history of selling computing services. Amazon launched Amazon Web Services in 2006, followed by Google Cloud in 2008 and Microsoft Azure in 2010. By 2015, AWS accounted for just 7% of Amazon’s revenue but 80% of its operating profit, giving the company a clear incentive to keep investing in a business that was already proving its worth. Meta, by contrast, was a high‑margin social media company with no reason to build a cloud offering. Now it is trying to leapfrog into a sector where rivals have decades of experience and established customer relationships, a position that puts Meta’s enormous capex programme under intense scrutiny.

Founder‑Run Firms and the Herding Instinct

Both Meta and Elon Musk’s businesses demonstrate how founder‑run companies can pivot rapidly on the whims of a CEO. The academic model of herd behaviour, as described in Abhijit Banerjee’s 1992 paper, shows how individuals rationally abandon their own information when they see others acting differently. Zuckerberg and Musk, once near the front of the queue with ground‑breaking ideas such as social networks and electric cars, are now six or seven spaces back in the line for data centres and chip manufacturing. Their decisions to follow the herd into unfamiliar territory amplify the risk that billions will be spent chasing a trend without the internal capabilities to turn that spending into sustainable profits.

What Investors Should Watch as Big Tech Bets on Compute

• Monitor Meta’s monetisation milestones: Investors should demand clarity on when and how the coding‑tools business and any external compute‑rental service will contribute to revenue. Without specific timelines, the negative free cash flow trajectory is unlikely to reverse soon.
• Scrutinise Tesla’s chip fab economics: The viability of an in‑house fabrication plant for a carmaker is unproven. Watch for public disclosures on cost per chip, capacity utilisation and any plans to serve third‑party customers; failure to articulate a credible model would signal a misallocation of capital.
• Assess core‑competency alignment: For SpaceX, investors should question whether the claimed synergies between rocket engineering and data‑centre operations are tangible or merely a justification for following the AI hype. Look for evidence of contracts or partnerships that convert the 86% capex allocation into paying customers.