Anthropic and the AI cluster effect in fundraising

The writer is a former global head of equity capital markets at Bank of America and is now a managing director at Seda Experts

When I was a child, my parents used to drive along Route 10 in East Hanover, past nondescript retail boxes and parking lots. It wasn’t postcard-scenic like the rest of New Jersey, but I was always struck by a stretch of mattress stores, one after another. All these years later, I still wonder: why cluster so close together? Why not spread out?

The economist Harold Hotelling explained why in 1929. Imagine two ice cream carts on a beach. Each could stand a quarter of the way in and serve half the space. Instead, they move towards the centre to poach each other’s customers, eventually ending up side-by-side. That’s also why you often see petrol stations grouped around an intersection or car dealerships bunched together at an off-ramp.

The AI giants appear to have enacted a capital-markets version of Hotelling’s Law. On Monday, Meta announced an up to $95bn equity raising — $35bn via concurrent offerings of common and preferred stock and up to $60bn of new non-voting stock through what is known as an “at-the-market” programme of selling equity over time.

The same day, Anthropic publicly filed the registration statement for its upcoming initial public offering. This is almost a repeat from last June when Alphabet announced an $80bn equity-raising plan around the time of SpaceX’s IPO, which raised $86bn. The big names keep parking their ice cream carts next to the newcomers.

This inverts equity capital markets orthodoxy. The conventional wisdom is to avoid launching two comparable deals at the same time, lest they cannibalise demand. Banks normally advise clients either to speed up preparations to get ahead of a competing deal or to wait for it to complete and investors to digest it.

So why not wait here? Anthropic’s IPO could raise up to $100bn, and investors will need to free up cash to subscribe, possibly by selling holdings in other AI-related names, such as the neoclouds, chipmakers, and hyperscalers. If you’re Meta, you don’t want to become a “source of funds” for this portfolio rebalancing. A stock offering brings Meta back into the conversation at a time when investors are juggling their AI allocations.

Meta’s initial offering consists of $22.5bn of common stock and $12.5bn of mandatory convertible preferred stock. The latter may appeal to a different set of buyers, including equity-income and convertible investors. By tapping more than one investor constituency, Meta can broaden its appeal even as it competes with Anthropic for capital.

The proposed $60bn at-the-market offering is also telling. In an ATM programme, the company dribbles stock into the market over time. Announcing its size now signals the prospect of substantial supply, giving investors another reason to think about Meta, even before Anthropic’s underwriters open the IPO order book.

The fundraisings reflect a change in capital market dynamics. The investment requirements of the AI build-out exceed the free cash flow of the large tech firms. Companies that have been buying back shares for years have now started to issue them. When several peers need equity at once, they are vying most of all for investor attention.

No issuer wants a rival to monopolise the news cycle, especially when investor appetite is uncertain. In recent weeks, AI stocks have traded off, IPOs of neoclouds have been postponed, and investors have reportedly pressed Anthropic to lower its valuation target. Waiting may mean allowing another issue to drain the pool of available capital and garner the attention that might otherwise be yours.

Recommended

There is, of course, another possible explanation. Maybe the deal clustering is a byproduct of an overheated market. Valuations remain high, the window is open, and companies are raising money while they can. On this reading, the rush of deals reflects naked opportunism as issuers try to catch the market before it turns.

It’s hard to distinguish herding from Hotelling. Simultaneous fundraising may reflect common circumstances rather than a conscious effort to pre-empt a rival. But the pattern suggests that issuers see the cost of being overlooked as a risk in its own right. There’s no point in waiting your turn if your turn might never come.

Does clustering benefit anyone? Retail strips attract footfall because shoppers can compare competing products in one place. If Anthropic’s debut generates more excitement around AI and draws in fresh money, then clustering is a win-win. If AI allocations stay fixed in the short run, the deals turn into a zero-sum contest, with both parties achieving less than they might have by waiting.

Hotelling’s ice cream carts profited from parking next to each other. The people on the beach, however, had to walk quite a bit further.