With 250,000 highly-desired Nvidia graphics processors, CoreWeave has become one of the most prominent “GPU clouds,” a status it hopes investors will value when it debuts on the public markets.
But the world of artificial intelligence hardware is moving so quickly that it raises questions about how long those chips will remain on the cutting edge and in demand. It’s a concern that could impact investor demand for shares of CoreWeave, one of the most anticipated IPOs in years.
CoreWeave, which rents out remote access to computers based on Nvidia AI chips,said in a financial filing this monththat most of its AI chips are from Nvidia’s Hopper generation. Those chips, such as the H100, were state-of-the-art in 2023 and 2024. They were scarce as AI companies bought or rented all the chips they could get in the wake of OpenAI ushering in the generative AI age with the release of ChatGPT in late 2022.
But these days, Nvidia CEO Jensen Huang says that his company’s Hopper chips are getting blown out of the water by their successors – the Blackwell generation of GPUs, which have been shipping since late 2024. Hopper chips are “fine” for some circumstances but “not many,” Huang joked at Nvidia’s GTC conference last week.
“In a reasoning model, Blackwell is 40 times the performance of Hopper. Straight up. Pretty amazing,” Huang said. “I said before that when Blackwell starts shipping in volume, you couldn’t give Hoppers away.”
That’s great for Nvidia, which needs to find ways to keep selling chips to the companies committed to the AI race, but it’s bad news for GPU clouds like CoreWeave. That’s because the New Jersey company models the future trajectory of its business based on how much it anticipates being able to rent Nvidia chips out for over the next five to six years.
Huang may have been kidding, but Nvidia spent much of its event detailing just how much better its Blackwell chips are. In Nvidia’s view, the best way to decrease the high cost of serving AI is by buying faster chips.
Blackwell systems are in full production and shipping to customers, and Nvidia plans to introduce an upgraded version of Blackwell in late 2026. When new chips come out, the older chips — the kind CoreWeave has a quarter of a million of — go down in price, Huang said. So too does the price of renting them.
Older chips don’t just stop working when new ones come out. Most companies, including CoreWeave, plan to use Hopper chips for six years. But Nvidia is telling customers that its newer, faster chips are capable of producing more AI content, which leads to more revenues at a better margin for clouds.
An H100 would have to be priced 65% lower per hour than an Nvidia Blackwell GB200 NVL system for the two systems to be competitive in price per output to a renter. Put another way, the H100 would have to rent at 98 cents per hour to match the price per output of a Blackwell rack system priced at $2.20 per hour per GPU, SemiAnalysis estimated, speaking generally about AI rentals.
H100s rented for as much as $8 per hour back in 2023 and often required long commitments and lead times, but now, usage of those chips can be summoned in minutes with a credit card. Some services now offer rented H100 access for under $2 per hour.
The industry could be entering a period where the useful life of AI chips is reduced, Barclays analyst Ross Sandler wrote in a note on Friday. He was focused on hyperscalers — Meta, Google and Amazon — but the trend affects smaller cloud providers like CoreWeave, too.
“These assets are becoming obsolete at a much more rapid pace given how much innovation and speed improvements happen with each generation,” Sandler wrote.
This threatens company earnings if they end up depreciating older equipment faster, he said.
CoreWeave says that if there were to be changes to the “significant” assumptions it makes about the useful lifetime of its AI infrastructure, it could hurt its business or future prospects. CoreWeave has also borrowed nearly $8 billion to buy Nvidia chips and build its data centers, sometimes using the GPUs it amassed as collateral.
Analysts and investors are also increasingly asking questions about the useful lifespan of these new AI systems and whether their financial depreciation schedules should be accelerated because the technology is improving so fast.
CoreWeave says in its filing that it seeks to offer state-of-the-art infrastructure and says it will continue spending to expand and improve its data centers.
“Part of this process entails cycling out outdated components of our infrastructure and replacing them with the latest technology available,” the New Jersey company said. “This requires us to make certain estimates with respect to the useful life of the components of our infrastructure and to maximize the value of the components of our infrastructure, including our GPUs, to the fullest extent possible.”
CoreWeave and Nvidia maintain a good relationship. CoreWeave will certainly buy more chips from Nvidia, which owns more than 5% of the New Jersey company.
“We’re super proud of them,” Huang said last week.
But Nvidia’s road map for releasing new chips that it proudly touts will make their predecessors obsolete is a threat to CoreWeave’s ambitions.
Amazon has restarted drone deliveries in two states after a months-long pause, the company confirmed.
In January, Amazon halted Prime Air deliveries in College Station, Texas, and Tolleson, Arizona, the two U.S. markets where it’s testing the service, as the company rolled out a software update to its drone fleet.
Amazon discovered an abnormality with the drone’s altitude sensor, caused by dust in the air, that could have caused its system to produce an inaccurate reading of its position relative to the ground, the company said. Amazon “never experienced an actual safety issue,” but said it opted to suspend deliveries while it corrected the issue.
The company brought drone deliveries back online last week after it completed the software update and received approval from the Federal Aviation Administration, Amazon spokesperson Av Zammit said in a statement.
“Safety underscores everything we do at Prime Air, which is why we paused our operations to conduct a software update on the MK30 drone,” Zammit said. “The updates are now complete and were approved by the FAA, allowing us to resume deliveries.”
An FAA spokesperson didn’t immediately provide a comment.
Zammit said Prime Air has seen “unprecedented levels of demand” since it resumed service. David Carbon, an executive who oversees Amazon’s drone program, wrote in a LinkedIn post last week that the company dropped a bottle of ZzzQuil sleep medicine at an Arizona customer’s home in “31 minutes and 30 seconds.” Carbon didn’t say how far the drone had to fly and Zammit declined to provide details.
For over a decade, Amazon has been working to bring to life founder Jeff Bezos’ vision of drones whizzing toothpaste, books and batteries to customers’ doorsteps in 30 minutes or less. But progress has been slow, as Prime Air has only been made available in the U.S. in College Station and Tolleson. A test site in Lockeford, California, was shuttered last April. The program was also hit with layoffs in 2023 as Amazon CEO Andy Jassy cut costs across the company.
The company also introduced a new version of its delivery drone, called the MK30, which is designed to be quieter than previous models and can fly in light rain.
Customers in College Station, a quiet suburban town that’s about 100 miles northwest of Houston, had previously complained about the drones’ noise levels. After rolling out the MK30, the company is also taking steps to relocate its drone hub farther away from residents’ homes later this year.
Before Amazon suspended drone deliveries, the MK30 crashed in two separate incidents during test flights at the company’s facility in Pendleton, Oregon. Last December, a software issue caused two drones to crash, according to Bloomberg. And in September, a pilot mistakenly caused a “mid-air collision” between two drones after he tested how the MK30 would perform when faced with a failed propeller, according to a federal crash report.
Another crash occurred on Feb. 21 during tests at the Pendleton site, which resulted in a drone sustaining substantial damage, according to a report compiled by the National Transportation Safety Board.
Amazon said the crashes were unrelated to its decision to halt drone operations. The company has said these kinds of incidents, which have also occurred with other models in previous years, are part of the testing process, as it pushes drone systems “up to the limits and beyond.”
LightSource cofounders: CTO Idan Mintz and CEO Spencer Penn
Courtesy: LightSource
With President Donald Trump set to impose sweeping tariffs on a wide swath of U.S. trading partners this week, corporate America is awash in uncertainty.
LightSource, a San Francisco startup whose software helps companies manage their procurement process, costs and vendor relationships, didn’t know what the president’s tariffs plan would look like before raising its first funding round. But the timing didn’t hurt.
LightSource has just closed a $33 million financing, led by Bain Capital Ventures and Lightspeed Venture Partners, with participation from J2 Ventures.
“Tariffs and trade winds are shifting so fast, it’s enough to make your head spin,” said Ajay Agrawal, a partner at Bain and now a board member at LightSource. “For a company with hundreds or thousands of different parts and suppliers — even just understanding what the impact will be on their whole enterprise is unbelievable.”
President Trump’s plans to slap “reciprocal tariffs” on all countries with duties on U.S. goods is set to be announced on Wednesday. Concerns surrounding the impact of those moves pushed the Nasdaq down more than 10% in the first quarter, the index’s biggest drop for any period since 2022.
Trump has already said he would impose 25% tariffs on “all cars that are not made in the United States.” Autos is a market that co-founder and CEO Spencer Penn knows well.
LightSource was started in 2021 by Penn and CTO Idan Mintz, while the two were working in different parts of Alphabet. Penn was at robotaxi unit Waymo, and Mintz was in the Google X “moonshot factory.”
Prior to Waymo, Penn worked at Tesla when the electric vehicle maker was starting to mass produce its popular Model 3 sedans. He said that finance, sourcing and engineering professionals have to work together to find, or sometimes custom order, high-quality parts. They also have to maintain their best supplier relationships while evaluating new potential vendors and negotiating fair prices.
Often these teams rely on “hundreds of disparate processes and information that’s stuck in thousands of emails, spreadsheets and randomly formatted invoices and contracts,” Penn said.
LightSource, which has about 30 employees, connects a company’s procurement-related information sources and systems to streamline that complex work. The aim is to speed up a company’s procurement process, saving the business time, money and pain while working with suppliers.
Mintz describes LightSource’s offering as a kind of “operating system” for procurement. Penn says it has the potential to do for procurement what Salesforce did for customer relationships.
Whether it’s a global pandemic, a natural disaster cutting off a shipping route, or a major shift in tariffs and trade policy, Mintz said, any supply chain disruption can make a huge difference to a company’s profit margins and its ability to deliver a product on time.
Current customers include consumer packaged goods companies, aerospace ventures, e-commerce companies and automotive giants.
Toronto , Canada – 20 June 2024; Andrew Feldman, co-founder and CEO of Cerebras Systems, speaks at the Collision conference in Toronto on June 20, 2024.
Artificial intelligence chip developer Cerebras said Monday that it has obtained clearance from a U.S. committee to sell shares to Group 42, a Microsoft-backed AI company based in the United Arab Emirates.
That clearance came from the Committee on Foreign Investment in the United States, or CFIUS, and it’s a key step for Cerebras in its effort to go public. Cerebras competes with Nvidia, whose graphics processing units are the industry’s choice for training and running AI models, but most of its revenue comes from a customer called Group 42.
Cerebras filed to go public in September but has not provided details on timing or size for the initial public offering. The regulatory overhang was tied to the company’s relationship with Group 42, which was the source of 87% of Cerebras’ revenue in the first half of 2024, made the IPO look uncertain.
“We thank @POTUS for making America the best place in the world to invest in cutting-edge #AI technology,” Andrew Feldman, Cerebras’ co-founder and CEO, wrote in a Monday LinkedIn post. “We thank G42’s leadership and the UAE’s leadership for their ongoing partnership and commitment to supporting U.S headquartered AI companies.”
Lawmakers have previously worried about Group 42’s connections to China. Last year Mike Gallagher, then a Republican member of Congress from Wisconsin, said in a statement that he was “glad to see G42 reduce its investment exposure to Chinese companies.” Microsoft later announced a $1.5 billion investment in Group 42.
Both Cerebras and Group 42 had given voluntary notice to CFIUS about the sale of voting shares, according to the Sunnyvale, California-based company’s IPO prospectus. Group 42 had agreed to buy $335 million worth of Cerebras shares by April 15, according to the prospectus. The two companies later changed the agreement to say Group 42 would be buying non-voting shares, prompting them to withdraw their notice, because they said they did not believe CFIUS had jurisdiction over sales of non-voting securities.
CFIUS did not immediately respond to a request for comment.
Just a handful of technology companies have gone public since 2021, as higher interest rates made unprofitable companies less desirable. But in recent months, Cerebras and a few technology-related companies have taken steps toward IPOs, and last week, AI infrastructure provider CoreWeavewent public.
CoreWeave shares fell 7% on Monday, its second day of trading.