The explosion of generative artificial intelligence (AI) has taken the world by storm, but one question all too rarely comes up: Who can afford it?
OpenAI Inc bled about US$540 million last year as it developed ChatGPT and said it needs US$100 billion to meet its ambitions, according to industry media The Information.
“We’re going to be the most capital-intensive start-up in Silicon Valley history,” OpenAI founder Sam Altman told a panel recently.
Photo: AFP
When Microsoft Corp, which poured billions of dollars in investment into OpenAI, was asked about how much its AI adventure would cost, the company answered with assurances that it is keeping an eye on its bottom line.
Building something even near the scale of what OpenAI, Microsoft or Google have on offer would require an eye-watering investment on state-of-the-art chips and recruiting prize-winning researchers.
“People don’t realize that to do a significant amount of AI things like ChatGPT takes huge amounts of processing power. And training those models can cost tens of millions of dollars,” independent analyst Jack Gold said.
“How many companies can actually afford to go out and buy 10,000 Nvidia H100 systems that go for tens of thousands of dollars a piece?” Gold asked.
The answer is pretty much no one and in tech, if you cannot build the infrastructure, you rent it and that is what companies already do massively by outsourcing their computing needs to Microsoft, Google and Amazon Web Services.
With the advent of generative AI, this dependency on cloud computing and tech giants deepens, leaving the same players in the driver’s seat, experts said.
The unpredictable costs of cloud computing “is a heavily underestimated problem for many companies,” Software AG chief product officer Stefan Sigg said.
Sigg compared cloud costs to electricity bills and said companies that do not know better are in for “a big surprise” if they let their engineers run up bills in the mad rush to build tech, including AI.
Microsoft’s signature cloud offer is Azure and some observers believe the giant’s all-in bet on AI is really about protecting Azure success and guaranteeing the cash cow’s future.
Azure has been the giant’s unsexy bread-winner for years, bringing in huge profits, but without attracting the headlines of an iPhone or social media that go straight to the consumer.
For Microsoft, “the golden goose is monetizing cloud with Azure, because we’re talking about what could be a US$20, US$30, US$40 billion opportunity annually down the road if the AI bet is successful,” Wedbush Securities Inc analyst Dan Ives said.
Microsoft CEO Satya Nadella said that generative AI is “moving fast in the right direction.”
Deeply respected on Wall Street, Nadella would have a six or nine-month grace period to show his bet is a winner, Ives said.
Microsoft acknowledges the risk, but insists that on AI, it must “lead this wave,” chief financial officer Amy Hood told analysts this month.
“We will charge for those AI capabilities, and then ultimately, we’ll deliver operating profit,” she said.
Piling up profit at the company founded by Bill Gates can only mean passing on the cost of AI to customers. From Main Street to Fortune 500, the dependency on the AI-amped would be an expensive one, and companies and investors are drumming up alternatives to at least reduce the bill.
“AI training, GPT training will become a very important cloud service going forward,” Spectro Cloud CEO Tenry Fu said.
His company, like many others in the sector, helps companies optimize cloud technology to reduce expenses.
“But after training, a company will be able to get their model back for real AI application” and the dependence on the cloud giants will hopefully be reduced, he added.
Regulators are hoping that they can keep up and not leave the giants in charge, imposing their terms on smaller companies.
“Law enforcers [must] ensure that ... opportunities and openings for competition ... are not getting squashed out by the incumbents,” US Federal Trade Commission Chairwoman Lina Khan told CNBC.
Intel Corp chief executive officer Lip-Bu Tan (陳立武) is expected to meet with Taiwanese suppliers next month in conjunction with the opening of the Computex Taipei trade show, supply chain sources said on Monday. The visit, the first for Tan to Taiwan since assuming his new post last month, would be aimed at enhancing Intel’s ties with suppliers in Taiwan as he attempts to help turn around the struggling US chipmaker, the sources said. Tan is to hold a banquet to celebrate Intel’s 40-year presence in Taiwan before Computex opens on May 20 and invite dozens of Taiwanese suppliers to exchange views
Application-specific integrated circuit designer Faraday Technology Corp (智原) yesterday said that although revenue this quarter would decline 30 percent from last quarter, it retained its full-year forecast of revenue growth of 100 percent. The company attributed the quarterly drop to a slowdown in customers’ production of chips using Faraday’s advanced packaging technology. The company is still confident about its revenue growth this year, given its strong “design-win” — or the projects it won to help customers design their chips, Faraday president Steve Wang (王國雍) told an online earnings conference. “The design-win this year is better than we expected. We believe we will win
Quanta Computer Inc (廣達) chairman Barry Lam (林百里) is expected to share his views about the artificial intelligence (AI) industry’s prospects during his speech at the company’s 37th anniversary ceremony, as AI servers have become a new growth engine for the equipment manufacturing service provider. Lam’s speech is much anticipated, as Quanta has risen as one of the world’s major AI server suppliers. The company reported a 30 percent year-on-year growth in consolidated revenue to NT$1.41 trillion (US$43.35 billion) last year, thanks to fast-growing demand for servers, especially those with AI capabilities. The company told investors in November last year that
Power supply and electronic components maker Delta Electronics Inc (台達電) yesterday said it plans to ship its new 1 megawatt charging systems for electric trucks and buses in the first half of next year at the earliest. The new charging piles, which deliver up to 1 megawatt of charging power, are designed for heavy-duty electric vehicles, and support a maximum current of 1,500 amperes and output of 1,250 volts, Delta said in a news release. “If everything goes smoothly, we could begin shipping those new charging systems as early as in the first half of next year,” a company official said. The new
RSS