Did the CHIPS Act Kill Intel?

Cue Monty Python peasant: “I’m not dead yet!”

Well, maybe not quite dead, but it doesn’t look good and the CHIPS Act, and the spirit thereof was certainly a big part of it.

Intel (NASDAQ: INTC) had placed a huge bet on its prospects of becoming the most important foundry (contract chip manufacturer) in the United States. It is not going to happen.

That honor has already been grabbed by Taiwan Semiconductor (NYSE: TSM) (TSMC). Its new Arizona plant is already manufacturing at the pace of 40,000 wafers in an elite node (4nm) per year, soon to climb to 100,000 and 200,000.

It is possible that TSMC would have built the Arizona plant regardless of the CHIPS act, yet the $11 billion in grants and loans it scored from CHIPS certainly did not slow or discourage TSMC’s move.

Now our colleague Dr. Robert Castellano points out that Samsung, TSMC’s leading rival, has just landed a $16.5 billion deal to make AI chips for Tesla (NASDAQ: TSLA)—in its Texas plants—in part thanks to $4.75 billion in CHIPS subsidies.

Meanwhile, Intel Foundry Services has lost essentially all its customers, shrinking from small billions to small millions in projected revenue.

Even before the CHIPS Act, U.S. anti-China sanctions had brought devastating retaliation against Intel. Plan A for getting Intel foundry services off the ground had been to acquire Tower Semiconductor (NASDAQ: TSEM), an Israeli leader in analog chips. The merger did not go forward because the Chinese blocked it in response to hostile U.S. actions against Chinese tech firms.

Meanwhile, on the design and device side, Intel suddenly faces a huge threat in its core CPU consumer product business. The personal computer market has kept Intel afloat for decades. A decade ago, Intel made both the chips for Apple and the majority of Windows PCs. Apple is long gone. Now even more ominously, Qualcomm (NASDAQ: QCOM) is rising as a fierce competitor for artificial intelligence ready CPUs.

Microsoft has already adopted Qualcomm’s Snapdragon chipset for all its Co-pilot devices, manufactured by Dell, HP, Samsung, Lenovo, ASUS, and more to come. The technology Qualcomm deployed to dominate the wireless communication market is the direct predecessor to the processing done by an AI neural network.

Playing catch up in this industry is extremely difficult. It is almost unheard of for a company once dominant in its field and in making a diverse group of products. Taiwan Semiconductor caught up and passed the entire world as a contract manufacturer of semiconductors. But it did that by extreme concentration in a new market by consciously riding the learning curve. TSMC took low margin projects, second sourcing and anything it could grab to ride the learning curve by repeated doublings of volume. Had it been a pre-existing firm such as Intel with a diverse set of obligations and under tremendous time and financial pressure, it’s far less likely that it could have succeeded.

Analysts were pleased that in its most recent quarter, Intel reported losing less money and taking in more revenue than expected. This is small beer. Intel has a miracle to pull off.

A first step is the announcement by CEO Lip-Bu Tan that Intel is essentially abandoning its plans to build up Intel Foundry Services with external customers. Ignoring multi-billion-dollar government bribes to stay in a business it no longer does well is a good first step as the Intel design side has remained strong.

In recent years, Intel has been a leader in improvements in architecture design, notably its EMIB (Embedded Multi-die Interconnect Bridge), the lower cost and more flexible alternative to silicon interposers for chiplet connectivity, and moving power delivery to the backside of the chip, to reduce routing congestion on the front of the chip, improve signal integrity, performance, and transistor density.

The integrated device manufacturer model, combining design and manufacturing, is exhausted for Intel for cutting-edge chips. Intel should leave manufacturing—and massive government subsidies—behind and refocus on what it does best.