Intel stops paying its dividend

Intel suspended its dividend with its second-quarter results, ending thirty-two years of uninterrupted payouts. It also announced a fifteen per cent cut to its workforce and lower capital spending. The shares fell twenty-six per cent in one session, their worst day in fifty years.

The warning signs were there. Intel had already cut its dividend by two-thirds in February last year, and in April it revealed that its manufacturing arm had lost $7 billion in 2023. The second-quarter numbers made things worse: a net loss of $1.6 billion on revenue of $12.8 billion, a shrinking gross margin and a weak forecast for the rest of the year. The company now promises $10 billion of cost savings next year, much of it from the job cuts.

A foundry Intel cannot afford

Pat Gelsinger's plan to turn Intel into a world-class contract chipmaker has collided with its finances. Leading-edge fabrication plants cost tens of billions of dollars up front, while Intel's older PC and server chip businesses are losing market share and cash flow to AMD and to ARM-based designs. With free cash flow gone, Intel could no longer pay a dividend and fund an unproven foundry at the same time.

How the plan was meant to work

When Mr Gelsinger returned as chief executive in 2021, he set out to reverse years of manufacturing slippage that had let Taiwan's TSMC take the lead in chipmaking technology. His strategy had two parts. Intel would race through five new manufacturing processes in four years to catch up, and it would open its factories to outside customers, competing with TSMC and Samsung as a foundry for other chip designers. Revenue from those customers would, in time, spread the enormous fixed cost of each new plant across more chips.

The flaw was timing. Fabs have to be built years before they fill with orders, and the bills arrive first. Intel has plants built, under way or planned in Arizona, Ohio, Ireland and Germany, while its internal customer, the part of Intel that designs PC and server chips, has been losing ground. Meanwhile the biggest growth market, chips for artificial intelligence, has been captured by Nvidia, whose designs are made by TSMC.

Help that did not close the gap

Intel has not been short of outside support. In March the American government awarded it up to $8.5 billion in grants under the CHIPS Act, plus access to loans, as part of an effort to bring leading-edge manufacturing back to America. It has sold minority stakes in its Arizona and Irish plants to Brookfield and Apollo, raising money without issuing shares. Microsoft has said it will use Intel's most advanced planned process for one of its own chips.

None of that has been enough. Subsidies arrive in stages and are tied to milestones. Selling stakes in factories brings in cash today but hands part of their future profits to someone else. A single large customer is a start, but a foundry needs many to fill its lines, and chip designers will not commit their best products until they see the process working at scale.

A precedent from AMD

Intel's own rival has been here before. In 2009 AMD, unable to fund both chip design and cutting-edge manufacturing, spun off its factories into a separate company that became GlobalFoundries. AMD then concentrated on design, outsourced its manufacturing to TSMC and, over the following decade, recovered to take share from Intel. The episode suggested that only a company with very deep pockets and a steady stream of customers can afford to run leading-edge fabs, and TSMC has since become that company.

Intel's management argues that the lesson cuts the other way: a Western alternative to TSMC is strategically valuable, and Intel is the only candidate. That may be true for America's national interest, but it does not follow that Intel's shareholders should bear the cost.

The case for holding on

Supporters of the strategy make a fair point. The spending is front-loaded, so free cash flow should improve once the heaviest construction is done. If Intel's next-generation process, known as 18A, performs as promised, the company could regain a technology lead it has not held for years and win outside orders. Cost cuts and lower capital spending buy time. Governments want an American foundry to succeed, which makes further support likely if things go wrong.

The trouble is that each of those supports depends on execution Intel has struggled to deliver. Its manufacturing roadmap has slipped before, and investors who were told for years that the turnaround was imminent have now lost their dividend as well.

What to watch

The tests are concrete. Announcements of large outside customers for 18A would do more for the shares than any cost-cutting target. Gross margins in the next few quarters will show whether the core business has stopped deteriorating. Any sign that the board is weighing a separation of the foundry, following the AMD model, would signal that the dual strategy has reached its limit.

Intel has chosen the foundry over its shareholders' income, and the market has marked it down accordingly. Unless the foundry wins big outside customers in the next couple of years, the shrinking legacy business probably cannot pay for it. Absent a major customer win by the end of next year, expect pressure to split or sell the manufacturing business, with shareholders who bought for income the clearest losers either way.