The capital and technology demands of remaining a world leader in the semiconductor business are becoming extreme. Just how extreme was evident from two announcements at Intel this week. They show how profoundly the company is adapting its traditional way of doing business as it tries to regain its lost technology edge and ensure the US is home to at least one world-leading chip manufacturer.
The first announcement was that it is selling a 49 per cent stake in two new manufacturing plants under development in Arizona to private equity firm Brookfield. This is an entirely new way to fund chip fabrication plants (fabs) as the cost of building the most advanced facilities soars. The Brookfield deal covers the first $30bn invested in Arizona, while Intel has put the long-term investment in its new fabs like the one it is building in Germany at more than $100bn each.
The deal is a byproduct of Intel’s decision to stand its ground on chip manufacturing, even as it struggles to regain the lead it has lost to TSMC and Samsung in the latest process technology.
Most other chipmakers have chosen a different route. Rival AMD threw in the towel a decade ago, quitting manufacturing to focus on design. That decision to specialise — while outsourcing manufacturing — has started to pay off, as AMD’s latest designs have eaten into Intel’s dominance in the market for the x86 chips used in most PCs and servers.
The vast scale that the latest chip fabs operate at is likely to exceed the needs generated by Intel’s in-house chip design business. That has meant entering the foundry market — making chips for other companies — in order to soak up the extra capacity.
The capital intensity this leads to is mind-boggling. Over the past decade, Intel’s capital spending averaged around 20 per cent of its revenue each year. In the future, it predicts that will rise to 25 per cent — and that’s before adding in new sources of money like government grants and co-investment arrangements like the Brookfield deal. All in, Intel indicated this could push capital expenditure up to more than 35 per cent of annual sales.
Exactly how the risks and rewards in the new financing arrangement are shared hasn’t been disclosed. But Intel hinted at protections it has offered, such as guaranteeing Brookfield a certain level of output at its new fabs, and said that the investment firm will earn a relatively fixed return on its 49 per cent stake, with some variability. In return for a financing cost that is higher than a straight borrowing arrangement, Intel believes it will keep most of the upside if the plant outperforms — meaning it will also retain most of the risk if it doesn’t.
The second sign of how drastically Intel’s business model is shifting came a day earlier, when chief executive Pat Gelsinger showed off a number of the company’s new chip designs. Rather than being based on a single piece of silicon, these are part of the company’s move to chips that combine several components, or “chiplets”, into a single semiconductor.
One advantage of so-called “disaggregated” chip designs like this is that Intel does not need to produce all of the parts itself. So if it fails to get back to the leading edge of manufacturing technology, it could turn to other companies to buy individual components it can’t make itself. Producing the other parts would still provide a way to keep its own new chip fabs humming.
This might hurt profit margins, since Intel would not be producing the most advanced components in its own fabs. On the other hand, designing and integrating these new processors might provide some margin protection and would put a premium on the chip packaging technology it has developed.
Whatever the outcome, this week’s developments point to a complex and technologically challenging new business model that is likely to take years to play out, as new fabs come online and new generations of chips are developed. It will also require a cultural transformation, as a company that was famous for its highly insular culture learns how to open up to technologies from other companies, while also trying to develop the new service mentality required to run a successful foundry business.
Making matters worse, Intel is trying to pull off this transformation in the face of eroding market share and a downturn in chip demand. Its shares have halved after the brief honeymoon period that followed Gelisinger’s appointment and its stock market value was eclipsed by AMD last month.
A new Intel may be starting to take shape, but there is a long slog ahead.