Former President Joe Biden’s CHIPS and Science Act already had flaws, but now his successor’s administration’s approach is complicating matters further.

“Your CHIPS Act is a horrible, horrible thing. We give hundreds of billions of dollars and it doesn’t mean a thing. They take our money and they don’t spend it. You should get rid of the CHIPS Act and whatever is left over, Mr Speaker, you should use it to reduce debt.” This is what President Donald Trump said in a joint address to Congress in March of this year.

The CHIPS and Science Act became law three years ago in August 2022. A proposed $52.7bn (£39.3bn) boost for the US technology sector, and semiconductors specifically, it originally enjoyed bipartisan support. But this has waned to the point where Trump can seriously tout its repeal.

The main political opposition today is Republican – but even there, senators and representatives whose states stand to benefit from CHIPS subsidies have been taken aback by Trump’s comments. Nevertheless, serious questions have now arisen outside Washington DC over whether CHIPS appropriately or effectively addresses the crisis that seeded it, as well as the US’s technological ambitions.

Three years in, the Act is stumbling against its own contradictions. It is a sprawling, costly programme designed to serve national security, industrial policy and electoral politics. However, it is struggling to deliver any of them coherently.

The Act had three interlinked thematic drivers:

1. The Covid-19-induced chip shortage that hit hardest in sectors such as automotive
2. Concern over Chinese access to and influence over technology needed for emerging dual-use technologies (chiefly AI)
3. Repatriating advanced chip manufacturing capacity to the US after its global share had fallen to 10%.

The current president’s preference for tariffs has since added a fourth factor: moving production to the US to avoid paying them.

The real shortages

The first big problem with the CHIPS Act is that it tackles its original goal as something of an afterthought.

The overarching goals of its $39bn manufacturing subsidy component are for the US to produce 20% of “leading-edge” logic chips by 2030, and to ramp domestic output of the high-bandwidth memory devices. Both support and extend AI platforms.

As a result, the lion’s share of an already earmarked $32.5bn in manufacturing grants has gone to around a dozen logic and memory bidders out of more than 150 companies that applied (there have been 32 recipients so far). The big winners include Taiwan Semiconductor Manufacturing Corporation (TSMC), Samsung, Intel, Micron Technology, SK Hynix, Texas Instruments, HP, BAE Systems and Microchip.

However, the cutting-edge technologies most of those bids foregrounded were not the ‘horseshoe nails’ of Covid. The most severe supply crunches hit parts produced on mature process nodes.

Shortages of older, reliable microcontrollers brought Detroit’s car manufacturers to a standstill, disrupted maintenance of military hardware and hampered repair of the medical ventilators that kept so many Covid in-patients alive.

The Center for Strategic and International Studies notes: “In early 2022, the US Department of Commerce released the results of a survey addressing the chip shortage and found that firms faced their most acute shortages not in cutting-edge chips but in legacy chips at the 40nm node or larger.”

There is a $10bn provision within CHIPS to address mature processes, and there have been some takers. But a deterrent is that older nodes deliver much lower margins. These margins only make commercial sense when maintaining an existing fab where the capital expenditure has already been absorbed. Most of that capacity is currently and therefore expected to remain in Asia, operated by companies such as TSMC and UMC in Taiwan and even SMIC in China (that, notably, operates under US sanctions).

China, China, China

This feeds into concerns over China. First, there is its influence over and attitude towards its neighbours, notably Taiwan. Second is its potential to disrupt sensitive Western supply chains (see also rare earths). Third are Western desires to keep the middle kingdom no more than halfway up the stairs when it comes to AI and its implications for national security.

In justifying CHIPS spending, a frequent political sleight-of-hand used by former President Joe Biden and his commerce secretary Gina Raimondo was to talk of companies bringing “advanced”, “more advanced” and sometimes “leading-edge” chip manufacturing to the US. One word they avoided, however, was “most”. There was good reason for that.

TSMC is today’s leader in not just foundry capacity but also foundry technology. It is also poised to be the main beneficiary of various forms of federal support.

Thanks to the CHIPS Act, the company’s Arizona expansion is expected to receive $6.6bn in grants, loans on advantageous terms of up to $5bn and a $25bn tax benefit against the Advanced Manufacturing Investment Credit (AMIC). A subsequent commitment to invest a further $100bn – announced by CEO CC Wei alongside Trump in March – is also set to qualify for a retained AMIC in Trump’s controversial ‘One Big Beautiful Bill’ measures. Carved out of CHIPS, it raises the credit from 25% to 35%.

Despite all this, TSMC will not be bringing its most advanced 2nm or later 1.4nm processes to Phoenix from day one – or days two, three, four, five and then some. Even more than the US fears the world’s most advanced chips being made on an island Beijing covets, Taiwan knows that this capability creates a ‘Silicon Shield’ – a near cast-iron guarantee that the US will intercede should China invade.

“This is not the very best that TSMC has to offer and it’s not going to be the very best. That is always going to be based [at its headquarter fabs] in Hsinchu,” says Jacob Feldgoise, senior data research analyst at Georgetown University’s Center for Security and Emerging Technology.

Though Intel may have been seen as a competitor and American national champion, a persistent manufacturing malaise continues to bedevil the company. Newly appointed CEO Lip-Bu Tan is very much in ‘corporate turnaround’ mode. TSMC’s planned extra $100bn investment in the US is widely thought to have been largely designed to – as Gary C Hufbauer, nonresident senior fellow at the Peterson Institute for International Economics, puts it – “scare the bejesus out of its competitors”.

Jobs – but at a cost

Yet, despite its weaknesses, forecasts show that the CHIPS Act is adding solid numbers of jobs – and jobs mean votes. They bolster the support it continues to receive from Democrats, Republicans in tight races and candidates on both sides in swing states (Arizona being the highest-profile example of the third cohort).

“Speaker [Mike] Johnson only has a thin majority [in the House of Representatives], so though he was also a sceptic [about the act] back in November, you haven’t seen much action since,” says Feldgoise. “When you look at where these awards have gone – well, I’d be very surprised at even a major revision, never mind repeal.”

The Semiconductor Industry Association (SIA) has – in conjunction with the Boston Consulting Group – been a major influence over the Act’s execution. In its latest report the SIA projects that the programme will create 122,000 construction jobs and 68,000 permanent jobs at fabs. These are up from earlier SIA forecasts of 93,000 in construction and 43,000 at fabs.

The questions here arise over the cost-per-job-created relative to typical salaries for the sector. In a detailed analysis, Hufbauer’s team at Peterson found that the original numbers suggested a subsidy-per-job of $185,000. By contrast, the average US salary range for a fab worker is $80-120,000 (and comparable numbers are a $60-90,000 range for TSMC Taiwan and a $45-70,000 range at Samsung’s Korean fabs). Even with the SIA increasing its core targets and suggesting a multiple against the fab jobs of 5.7x, the Act still looks expensive.

“And it is,” adds Hufbauer, “but once you’ve effectively given states things like that, it’s politically unwise to take them away.”

Running on bureaucratic fumes

Then again, as the late, great writer Tom Wolfe once observed, you can render any bureaucratic process “arteriosclerotic”.

“I would say that the main force multiplier for the government is the technology, though the jobs are politically important,” says Mallory Block, public policy analyst at The Conference Board. “But there is an issue around how funds through the act will be disbursed.” Hufbauer and Feldgoise share this concern.

CHIPS is complex legislation, aims to serve multiple masters and conflicts with the current President’s instincts both economically and – given it was a ‘win’ for Biden – personally.

The CHIPS Program Office (CPO), which originally oversaw the manufacturing side of the initiative, had approximately 150 staff by the end of Biden’s presidency. By March, rumour had it that the CPO had fallen prey to the mercurial Department of Government Efficiency – 20 staff were told to retire; 40 on ‘trial’ contracts were sacked.

Among CPO’s leadership, both COO Sara Slayton O’Rourke and chief investment officer Todd Fisher have left. Its overall director Michael Schmidt joined the Micro Advisory Partners Consultancy as a senior advisor in June.

There have been rumours that staffing may now be as low as 30.

In that context, Block provides an idea of the tricky work still to be done. “Overall, you had just over $32.5bn in grant awards and nearly $6bn in loans to 32 companies across 48 projects, which had only just started to be disbursed when Biden left office.

“What’s important to remember here is that this money first got awarded to companies through a preliminary memorandum of terms or PMT. That’s a non-binding agreement between the companies and [the Department of] Commerce, and there is due diligence and other processes that follow the PMT. This can still be ongoing.”

This has implications beyond the flow of funding. “Uncertainty there can make companies more hesitant about moving forward with projects because they are now wondering how much they will need to absorb if funding is somehow lost,” Block adds.

Unintended consequences

The hindsight on this mess – albeit one that does seem to be muddling along – also has it that the US should have coordinated its programme with allies. Instead, it has inspired them to compete and, more damaging, potentially duplicate Washington’s efforts. It was ‘America First’ before the return of Mr America First himself, but, ironically, he doesn’t care much for this version.

The bias towards manufacturing subsidies means that instead there is now five-way competition over incentives between the US, EU, Japan, Taiwan and South Korea based on a mix of subsidies and tax credits globally worth roughly three times what the CHIPS Act originally tabled at around $150bn (or 15 megafabs targeting just two or three nodes). There is a growing threat of massive overcapacity.

Meanwhile, more than just CHIPS funding is stuttering. Alongside forecasts of 68,000 new US fab jobs, other industry data implies an accelerating shortfall in sufficiently skilled employees of 67,000. There is also a shortage of suitably qualified construction capable of working to the most stringent clean room standards – a new TSMC partner fab in Japan was completed by Asian contractors in two years, while a similar project in the US, also using local labour, took significantly longer.

And, as Feldgoise notes, the programme may only now be moving into its most critical phase: R&D at the CHIPS-funded National Semiconductor Technology Center, which formally opened in Albany, New York, on 14 July. What is the US’s roadmap to 1.4nm and to quantum, and what may lie beyond even that?

Trump is probably right that there is a fair tranche of ‘horrible’ in the Act. It is a bold experiment in reshoring and tech security. But it is also proving a confused, costly compromise that lacks strategic clarity, bureaucratic stability and global alignment. That hopefully will favour moves towards a fix rather than scorching the earth.