chips

October 1, 2025 0 comments

Jensen Huang sat down, speaking as part of the BG2 podcast.

Jensen Huang says China is ‘nanoseconds behind’ the US in chipmaking, calls for reducing US export restrictions on Nvidia’s AI chips

by admin September 28, 2025

Nvidia CEO Jensen Huang says China is just “nanoseconds behind” the U.S. in chipmaking and that Washington should stop trying to wall off the market. Speaking on the BG2 podcast, Huang argued that allowing companies like Nvidia to sell into China would serve American interests by spreading U.S. technology and extending its geopolitical influence. “We’re up against a formidable, innovative, hungry, fast-moving, underregulated [competitor],” Huang said, talking about the pedigree of China’s engineers and controversial 9-9-6 working culture.

His comments come as Nvidia hopes to ship its H20 AI GPU to Chinese customers again, following a months-long pause tied to new U.S. export rules. The Commerce Department is understood to have begun issuing licenses for the H20 in August, and Nvidia is already working on a successor chip designed to comply with current restrictions while offering better performance. The company has not confirmed specs, but it would be Nvidia’s second attempt to tailor an AI accelerator specifically for the Chinese market since the original A100 and H100 bans took effect.

NVIDIA: OpenAI, Future of Compute, and the American Dream | BG2 w/ Bill Gurley and Brad Gerstner – YouTube

Watch On

China, meanwhile, is accelerating its own plans to become self-sufficient. Huawei’s new Atlas 900 A3 SuperPoD systems, powered by the company’s Ascend 910B chips, are now shipping in volume. The company has laid out an ambitious roadmap through 2027 with next-gen Ascend silicon that aims to match or exceed current-gen performance. These systems are CUDA-free by design and optimized for Chinese-built software stacks, a shift that puts real pressure on Nvidia’s dominance, which, according to Huang, previously held a 95% market share in China.

You may like

Chinese hyperscalers are backing that roadmap with capital. Baidu, Alibaba, Tencent, and ByteDance are all investing in custom silicon, either through internal chip teams or by funding startups. That includes firms like Tencent, which has announced it has fully adapted its infrastructure to support homegrown silicon. Asked what he sees in the near future, Huang said, “They [China] publicly say… they want China to be an open market, they want… companies to come to China and compete in the marketplace… and I believe and I hope that we return to that.”

Nvidia’s approach to that is to maintain a foothold in China and play both sides of the geopolitical divide. The H20 may be hobbled compared to the company’s leading chips, but it gives Chinese companies a path to stay within the Nvidia ecosystem — at least for now.

Follow Tom’s Hardware on Google News, or add us as a preferred source, to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button!

September 28, 2025 0 comments

Enfabrica’s ACF-S and EMFASYS architecture could change how AI clusters process tens of thousands of chips efficiently

by admin September 26, 2025

Nvidia’s acquisition brings Enfabrica engineers directly into its AI ecosystem
EMFASYS chassis pools up to 18TB of memory for GPU clusters
Elastic memory fabric frees HBM for time-sensitive AI tasks efficiently

Nvidia’s decision to spend more than $900 million on Enfabrica was something of a surprise, especially as it came alongside a separate $5 billion investment in Intel.

According to ServeTheHome, “Enfabrica has the coolest technology,” likely because of its unique approach to solving one of AI’s largest scaling problems: tying tens of thousands of computing chips together so they can operate as a single system without wasting resources.

This deal suggests Nvidia believes solving interconnect bottlenecks is just as critical as securing chip production capacity.

You may like

A unique approach to data fabrics

Enfabrica’s Accelerated Compute Fabric Switch (ACF-S) architecture was built with PCIe lanes on one side and high-speed networking on the other.

Its ACF-S “Millennium” device is a 3.2Tbps network chip with 128 PCIe lanes that can connect GPUs, NICs, and other devices while maintaining flexibility.

The company’s design allows data to move between ports or across the chip with minimal latency, bridging Ethernet and PCIe/CXL technologies.

For AI clusters, this means higher use and fewer idle GPUs waiting for data, which translates into better return on investment for costly hardware.

Another piece of Enfabrica’s offering is its EMFASYS chassis, which uses CXL controllers to pool up to 18TB of memory for GPU clusters.

This elastic memory fabric allows GPUs to offload data from their limited HBM memory into shared storage across the network.

By freeing up HBM for time-critical tasks, operators can reduce token processing costs.

You may like

Enfabrica said reductions could reach up to 50% and allow inference workloads to scale without overbuilding local memory capacity.

For large language models and other AI workloads, such capabilities could become essential.

The ACF-S chip also offers high-radix multipath redundancy. Instead of a few massive 800Gbps links, operators can use 32 smaller 100Gbps connections.

If a switch fails, only about 3% of bandwidth is lost, rather than a large portion of the network going offline.

This approach could improve cluster reliability at scale, but it also increases complexity in network design.

The deal brings Enfabrica’s engineering team, including CEO Rochan Sankar, directly into Nvidia, rather than leaving such innovation to rivals like AMD or Broadcom.

While Nvidia’s Intel stake ensures manufacturing capacity, this acquisition directly addresses scaling limits in AI data centers.

September 26, 2025 0 comments

WASHINGTON, DC - JUNE 9: A U.S. Department of Commerce sign is displayed at the Herbert C. Hoover Federal Building on June 9, 2025 in Washington, DC. (Photo by Kevin Carter/Getty Images)

Trump administration is reportedly planning to tariff US tech firms that don’t source equal numbers of imported and American chips

by admin September 26, 2025

Every tech firm in the US heavily relies on the likes of China and Taiwan for its products, whether it involves the wholesale manufacturing of them or the supply of the vast number of semiconductor chips and components required. However, if a purported idea being considered by the Trump administration comes to fruition, they will all need to massively reduce imports and switch to locally-made chips to avoid being hit with a fresh tariff.

That’s according to a report by the Wall Street Journal, which claims that Commerce Secretary Howard Lutnick has already mooted the idea with various executives within America’s semiconductor industry. If we use Nvidia as an example, it currently relies almost exclusively on companies outside of the US for all the chips and other electronic components that are used to manufacture its graphics cards and AI data servers.

Its GPUs and CPUs are made by TSMC in Taiwan, with circuit boards and the host of parts that are fitted to them produced in China. Nvidia tends to use Micron for VRAM chips more than any other firm, and while that company is US-based, it also has production facilities in Singapore and China.

To comply with a mandate that requires it to maintain a 1:1 ratio of locally-produced semiconductor chips versus those that it imports, Nvidia would need to drastically change its supply chain somehow. Either that, or it would have to rely on the majority of its suppliers having facilities within the US to produce said components.

At the moment, there’s no indication of the nature or size of the tariff that would be applied if companies failed to reach the ratio target, but even if the threat of it is big enough to make all US tech companies immediately comply, one question remains unanswered. And it’s because there is no answer for it.

TSMC’s chip foundry in Arizona. America’s going to need a lot more of these. (Image credit: TSMC)

How is America’s semiconductor industry supposed to match the combined output, breadth of products, and level of technological accomplishment of Taiwan, South Korea, China, Japan, and Singapore? Despite having the likes of Intel, GlobalFoundries, and Micron, as well as fabrication plants from Samsung and TSMC, the supply chain for the global tech market is predominantly based outside of the US.

If one assumes that it can be scaled up to the level required to meet the 1:1 demand, it certainly can’t happen overnight, and the cost for adjusting the supply chain to this extent is likely to be enormous. So much so that it’s possible that any tariff would pale in comparison.

WSJ’s report also claims that the plan would allow companies to make manufacturing pledges, to give themselves sufficient time to build the required infrastructure in America, without incurring the tariff. There may also be a relief period if and when the plan is introduced, to allow for US-based production to be ramped up.

While it can be argued that having a more equally distributed semiconductor supply chain is beneficial for stability and security reasons, the economic impact of forcing it to significantly adjust so rapidly could be too much for the industry to bear; at the very least, tech companies that are currently struggling with uncertain revenues or low profit margins would not welcome the plan.

For the US tech industry, this could ultimately be good news or catastrophic news, but until any official statement is made by the Trump administration, we’re just left with speculation. Any move to significantly reduce chip imports might seem like a great idea, but with the devil being in the details, and details being thin on the ground right now, tech firms are probably feeling a tad jittery about all of this.

Best PC build 2025

All our favorite gear

September 26, 2025 0 comments

Nvidia pours $100 billion into OpenAI and supplies millions of chips, raising fresh questions about competition and market concentration

by admin September 25, 2025

Nvidia commits $100 billion to OpenAI while reinforcing demand for its hardware
Partnership builds massive data centers and fuels concerns over circular investment structures
Analysts warn deal may raise antitrust scrutiny as Nvidia strengthens AI dominance

Following its recent surprise $5 billion Intel deal, Nvidia is spending big again, this time committing up to $100 billion to OpenAI alongside supplying millions of its chips.

The move fits a broader pattern in which Nvidia channels money into businesses that rely on its own hardware, from $6.3 billion in CoreWeave to $700 million in nScale, effectively reinforcing demand for its products while bypassing hyperscalers like Google and Microsoft which are racing to reduce their dependence on Nvidia’s hardware.

This latest investment into the world’s best-known AI firm immediately lifted Nvidia’s market value by more than $220 billion.

You may like

Circular structure

The deal involves a circular structure and will see Nvidia will buy non-voting shares in OpenAI, which OpenAI will then spend mostly on Nvidia systems.

Citing people familiar with the matter, Reuters says the partnership will begin with a $10 billion investment and scale as OpenAI deploys more computing power.

“This is the biggest AI infrastructure project in history,” Nvidia founder and CEO Jensen Huang said in an interview with CNBC’s Jon Fortt. “This partnership is about building an AI infrastructure that enables AI to go from the labs into the world.”

He said the companies will build data centers capable of running next-generation AI models, powered by Nvidia’s new Vera Rubin platform.

The first data centers are due online in 2026 and require 10 gigawatts of power, roughly equal to the needs of 8 million US households.

OpenAI chief executive Sam Altman said the capacity was essential for the company’s ambitions.

“Building this infrastructure is critical to everything we want to do,” Altman said. “This is the fuel that we need to drive improvement, drive better models, drive revenue, drive everything.”

You may like

Analysts welcomed the long-term demand for Nvidia’s products but warned about the structure of the deal.

“On the one hand this helps OpenAI deliver on some very aspirational goals for compute infrastructure,” said Stacy Rasgon of Bernstein. “On the other hand the ‘circular’ concerns have been raised in the past, and this will fuel them further.”

Kim Forrest, Chief Investment Officer, Bokeh Capital also sounded a note of caution. “This sounds like Nvidia is investing in its largest customer. These arrangements can be beneficial for both parties. But there can be dangers as well. Being totally linked with each other can cause for short-sightedness and can make an entry point for other chip competitors to come into other AI companies and woo them,” she said.

MarketScreener quotes Rebecca Haw Allensworth, an antitrust professor at Vanderbilt Law School, who says there are concerns that Nvidia could favor OpenAI with better pricing or faster delivery times.

“They’re financially interested in each other’s success,” she said. “That creates an incentive for Nvidia to not sell chips to, or not sell chips on the same terms to, other competitors of OpenAI.”

An Nvidia spokesperson denied this would be case, saying, “We will continue to make every customer a top priority, with or without any equity stake.”

Nvidia plans to invest up to $100 billion in OpenAI as part of data center buildout – YouTube

Watch On

September 25, 2025 0 comments

Qualcomm Debuts Snapdragon X2 Elite and X2 Elite Extreme, Its Next-Gen Laptop Chips

by admin September 24, 2025

Perhaps the most exciting part of the chip is the graphics performance: The Snapdragon X2 Elite and X2 Elite Extreme both use a new Adreno GPU architecture. Qualcomm says power efficiency has improved by 2.3X, so hopefully that means these laptops can scale up performance when it’s needed, in games or creative applications. Qualcomm says gaming will take a big step forward in this next generation. Qualcomm showed a chart with 2.2X faster frame rates in Hitman World of Assassination and 2.1X faster in Black Myth Wukong. Of course, these Qualcomm laptops continue to rely exclusively on integrated graphics, meaning the discrete GPUs in dedicated gaming laptops are still on another level of performance. Snapdragon X2 chips will also support higher refresh rate screens, now up to 144 Hz and up to three external 5K monitors.

Lastly, the company claims there’s a much more powerful neural processing unit (NPU) with 80 TOPS (trillions of operations per second) of on-device AI processing performance. The Snapdragon X set the initial bar with 50 TOPS, which Apple, Intel, and AMD then had to match in their next chips. We’ve yet to see the explosion of on-device AI processing be relevant for the average person, but hopefully, more performance will provide an incentive for developers to build more unique AI experiences that can utilize the NPU.

Qualcomm still boasts “multi-day” battery life on these new chips, though performance was the priority in the company’s product presentation. This is a change from the original marketing around the first-generation Snapdragon X chips, which centered battery life as its primary selling point.

As always, it helps to go first. These second-generation PC chips have been announced months ahead of devices from competitors, which may not arrive until the first quarter of 2026, making it easier to compare with previous-gen chips. Intel is expected to launch its highly anticipated A18 chips (also known as Panther Lake) later this year, and Nvidia’s $5 billion investment in Intel could change everything in the world of PCs. Apple is also expected to debut an M5 chip next month.

A New Smartphone Chip

Alongside new PC chips, Qualcomm also detailed its recently announced mobile chips, the Snapdragon 8 Elite Gen 5. These were officially unveiled last week, but at the company’s Snapdragon Summit conference, it shared new details on how these chips will perform.

Barely a week has gone by since Apple’s A19-powered iPhones hit the market, but Qualcomm is already claiming that its Snapdragon 8 Elite Gen 5 is the “fastest mobile CPU in the world,” with a record clock speed of 4.6 GHz (Apple’s A19 Pro has a clock speed of 4.26 GHz). Like the PC chips, graphics and AI processing have the largest leaps in performance, with a purported 23 percent faster GPU and 37 percent faster NPU. Some slightly faster single-core performance in your next phone might not be noticeable, but the significantly faster GPU could have a bigger impact on mobile gaming.

September 24, 2025 0 comments

An image of an experimental cooling solution developed by Microsoft, that uses microfluidics to get coolant directly into the processor's silicon.

Microsoft is resorting to laser etching AI-designed cooling channels directly into data center chips to tame their massive heat

by admin September 24, 2025

Introducing microfluidic cooling: a breakthrough in chip cooling technology – YouTube

Watch On

If you think the power consumption of today’s gaming graphics cards is bad, it’s nothing compared to how energy the massive processors in AI and data systems use. All that power ends up as heat, resulting in chip cooling being a serious challenge. Microsoft reckons it has a great solution, though, and it’s all about getting water into the processors themselves.

The most complex direct-die, liquid cooling loops you’ll see in a gaming PC all involve using a chamber that mounts on top of the CPU. At no point does the coolant ever touch the chip directly. In a recently published blog, Microsoft explains how it has developed a system that does precisely that.

By etching the surface of the processor die with an intricate pattern of tiny channels, water can then be pumped directly into the silicon itself, albeit to a very shallow depth.

The keyword to describe this is microfluidics, a technology that’s been around for many decades, and if the history of consumer tech is anything to go by, it’ll be a phrase plastered across every CPU cooler within a couple of years (though not actually do anything).

This might all just seem like Microsoft is cutting a few grooves into the chip and having water to flow through it, but it’s far more complicated than that. For a start, the channels themselves are no wider than a human hair, and they’re not just simple lines either. Microsoft employed the services of Swiss firm Corintis, which used AI to determine the best pattern for maximum heat transfer.

(Image credit: Microsoft)

The end result is a network of microchannels that genuinely look organic, though at first glance, you’d be forgiven for thinking the complex patterns were just manufacturing defects. It certainly looks super cool (pun very much intended).

Microsoft claims the tech is up to three times more effective at removing heat from a massive AI GPU than a traditional cold plate (aka waterblock), citing a 65% reduction in the maximum temperature rise of the silicon.

Since all the coolant transfer apparatus doesn’t need to be right on top of the microchannels, the system can also be applied to stacked chips, with each one etched before mounting. This way, each die within the stack is cooled individually, meaning they can operate closer to their maximum specifications than with a normal cold plate.

Take AMD’s X3D processors, for example. These all have one stacked chip underneath the heatsink: a Core Complex Die (CCD) bonded to a 3D V-Cache die. Each one acts as a thermal barrier to the other, though the CCD does generate much more heat than the cache die. If these could be both cooled via microfluidics, you’d be able to operate them both at higher clock speeds.

Of course, such complex tech isn’t cheap to develop or implement, and the likelihood of it ever appearing at the consumer level is very slim. But I wouldn’t be surprised if somebody takes an RTX 5090, rips off the heatsink, and swaps it for a homebrewed microfluidic cooler.

There again, if ramping up power consumption is the only way AMD, Intel, and Nvidia can keep improving chip performance, perhaps we might see etched processors and direct-die cooling being standard fare in our gaming PCs. After all, it wasn’t that long ago when heatpipes and vapour chambers were phrases never to be uttered by a PC component manufacturer, but now they’re in coolers of every kind.

Best CPU coolers 2025

All our current recommendations

September 24, 2025 0 comments

OpenAI is reported as Broadcom’s fourth XPU customer, joining Google, Meta and ByteDance in designing chips to reduce reliance on Nvidia

by admin September 11, 2025

OpenAI (probably) joins Google, Meta and ByteDance in Broadcom’s custom ASIC partnership
Broadcom secures $10 billion AI rack orders as Nvidia faces new rivals
Nvidia’s largest customers pursue in-house chips with Broadcom guiding the transition

As we’ve reported more than a new times in the past, the AI hardware market is changing, with some of Nvidia’s biggest customers looking for ways to cut costs and gain more control over their systems.

Rather than relying solely on Nvidia’s costly GPUs, companies are beginning to design their own ASICs tailored to their workloads.

Broadcom is one of the bigger players in this space, offering the expertise needed to turn those custom designs into production-ready chips and systems.

You may like

Reporting on Broadcom’s financial results for its third quarter, The Next Platform says the silicon supplier has now secured a fourth customer for its custom XPU program, adding to partnerships with Google, Meta, and ByteDance.

Industry reports and timing, suggest this newest client is OpenAI, which is developing its own inference processor known as Titan under the leadership of Richard Ho, a former Google TPU engineer.

Nvidia still dominates the market of course – by some way – with its Blackwell GB300 NVL72, but deploying such rackscale systems is expensive, and firms with massive AI models want hardware designed to better match their needs.

Custom ASICs are seen as a way to rein in costs while offering greater flexibility than an off-the-shelf GPU and Broadcom is well positioned to guide complex accelerator projects through design, production, and packaging.

On a call with Wall Street analysts, chief executive Hock Tan expanded on Broadcom’s unnamed fourth client (cough, OpenAI, cough), saying, “Now further to these three customers, as we had previously mentioned, we have been working with other prospects on their own AI accelerators.”

“Last quarter, one of these prospects released production orders to Broadcom, and we have accordingly characterized them as a qualified customer for XPUs and, in fact, have secured over $10 billion of orders of AI racks based on our XPUs,” Tan continued.

“And reflecting this, we now expect the outlook for our fiscal 2026 AI revenue to improve significantly from what we had indicated last quarter.”

That $10 billion figure refers to complete AI rack systems, not Broadcom’s share for the underlying chip design.

Revenue from those orders is scheduled to begin in the third quarter of fiscal 2026.

It’s clear that Nvidia’s biggest buyers are no longer content to depend solely on GPUs, and by investing in ASICs they are betting that custom hardware will bring efficiency and control. With its expertise, Broadcom is positioning itself as the company that can make those designs a reality.

September 11, 2025 0 comments

Intelligence everywhere: scaling the IoT with flexible chips

by admin August 27, 2025

From smart watches to smart supply chains, digital transformation is reshaping how we live, work and interact. But the true vision of an interconnected world – where billions of everyday items are embedded with intelligence – remains just out of reach.

That’s not because the ambition is too bold. In fact, it’s widely recognized. McKinsey estimates the Internet of Things (IoT) could deliver up to $12.5 trillion in global value by 2030. Extend that connectivity to low-cost everyday objects, and the potential increases dramatically.

So why hasn’t it happened yet?

You may like

Shane Geary

Social Links Navigation

Executive Vice President of Manufacturing and Operations at Pragmatic Semiconductor.

The biggest roadblock is supply. A truly connected world needs a colossal volume of chips. But we’re not talking about the advanced processors that power smartphones or autonomous vehicles – we’re talking about simpler, highly distributed, low-cost chips that can be embedded into everyday items.

These so-called ‘legacy’ chips – typically manufactured on 28nm or larger nodes – are anything but outdated. They’re critical to modern electronics, supporting high-performance processors and underpinning countless devices across consumer, industrial and automotive markets.

We saw just how vital they are during the COVID-19 pandemic. Shortages of these chips brought global manufacturing to a standstill, delaying everything from cars to home electronics. The ripple effects were felt across entire economies. As the number of smart devices scales, how to prevent a reoccurrence is an ever-present consideration.

So why not just build more chip factories?

Traditional fabs can’t meet future needs

Scaling legacy chip production is incredibly difficult. The equipment is aging and hard to replace. New fabs cost tens of billions and take years to build. And since mature-node chips don’t deliver the same profit margins as cutting-edge silicon, there’s little financial incentive to invest.

Even worse, legacy fabs are environmentally demanding. They consume vast amounts of energy and water. While next-gen fabs are built for sustainability, retrofitting legacy fabs to reduce their carbon footprint is expensive and complex.

To unlock true scale, we need a fundamentally different approach: one that’s faster, affordable, more flexible – and designed for sustainable, high-volume production.

Enter FlexICs

Flexible integrated circuits (FlexICs) are a new class of semiconductor: a chip that’s ultra-thin, physically flexible, and radically more sustainable.

Instead of traditional silicon wafers and high-temperature processing, FlexICs use thin-film technology and a low-energy manufacturing process. Think: less water, less energy, fewer harmful chemicals – and much faster turnaround.

This breakthrough unlocks a faster, more agile development cycle. Designs go from tape-out to volume production in weeks, not months. Costs are dramatically reduced. And it becomes viable to prototype, iterate and customize chips without the high stakes typically associated with silicon design.

Built for ubiquitous intelligence

This isn’t just a manufacturing story. It’s a transformation in how and where intelligence can be deployed.

FlexICs enable connectivity in places where traditional chips simply can’t go. Their ultra-thin, flexible form factor means they can be embedded directly into products or packaging, even on curved surfaces, delivering smarter, more connected experiences almost anywhere.

And they support standard communication protocols such as NFC, unlocking seamless, item-level intelligence for mass-market products – and the fastest digital connection between brands and their customers.

This embedded intelligence has multiple roles to play, from driving loyalty programs and exclusive content based on location or season, to product authentication, provenance information – or even facilitating improved reuse and recycling for a more circular economy.

It also offers a practical route for embedding real-time data collection into previously passive environments, whether that’s logistics chains, recycling systems, or agricultural operations.

Wherever you need intelligence, FlexICs bring it – sustainably, affordably, and at scale.

The future is flexible

But its goal isn’t about competing with silicon fabs. It’s about complementing them – addressing the unmet need for scalable, cost-effective, environmentally conscious production of the chips that will power the next wave of smart, connected systems.

We’re standing at the threshold of a hyperconnected future. But realizing that future depends on embedding intelligence at unprecedented scale. Flexible chips are the key to unlock that potential and make the vision of an interconnected world a reality.

We list the best performance management software.

This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro

August 27, 2025 0 comments

Trump Is Betting Big on Intel. Will the Chips Fall His Way?

by admin August 21, 2025

The US government is aiming to take an equity stake in Intel in exchange for grants the company was already committed to receive under the Biden era CHIPS Act, according to comments US commerce secretary Howard Lutnick made in an interview with CNBC. The move is part of the government’s efforts to boost US chip manufacturing.

“We should get an equity stake for our money, so we’ll deliver the money which was already committed under the Biden administration,” Lutnick said. “We’ll get equity in return for it.” Previously, the government was discussing taking a 10 percent stake in Intel, according to the New York Times.

The deal could help the venerable chipmaker fund its US-based semiconductor fabrication plants, or fabs, which have required billions of dollars to construct and maintain, even as demand for Intel chips has waned in recent years. Some chip industry experts and members of the Trump administration say that keeping Intel afloat is essential to US national security, because it lessens the country’s reliance on chipmakers overseas.

But analysts and one notable economist say a potential tie-up between Intel and the US government could present a conflict of interest and may not result in the kind of domestic chipmaking industry the administration is angling for.

“It’s not the right policy to have the US government own things, to have privatization in reverse,” says Stephen Moore, a visiting fellow at The Heritage Foundation and a former senior economic adviser to Trump’s 2016 campaign. “That’s similar to Europe’s industrial model, and we haven’t done that often here in the US, because a lot of it ends up failing.”

Government Intervention

The US government has some history of investing in the private sector. Moore cites a 1980s program called the Synthetic Fuels Corporation, a federally directed multibillion-dollar investment in companies producing liquid fuels from coal, oil shale, and tar sands. It was hailed by President Jimmy Carter as “the cornerstone of our energy policy” and had fallen apart by 1986.

Then, in the wake of the 2008 financial crisis, the US government stepped in with multibillion-dollar bailouts to stop US automakers and some banks from going under. Those funds were issued either through the Troubled Asset Relief Program, in which the US Treasury Department bought up or guaranteed toxic assets, or in the form of bridge loans. Many were eventually repaid.

More recently, the Department of Defense agreed to fund a US-based rare-earth magnet company, MP Materials, via equity and loans, in order to expand production and decrease the country’s reliance on China. The deal would in theory give MP Materials the capital to increase its manufacturing capacity from 3,000 to 10,000 metric tons.

Moore says the ideal scenario is that these arrangements between the government and private industry have an end point. “It should be an agreement to own a short-term stake and then divest,” he says.

But the current Trump administration has been taking some of these public-private business dealings a step further: In June, the administration approved a partnership between Japanese steel company Nippon Steel and Pittsburgh-based US Steel, dependent on a national security agreement and a so-called golden share provision. The government insisted that it have a say in US Steel’s company decisions, including board appointees and future relocation plans. (This deal was also designed to help the US compete with China on steel production.)