
Follow ZDNET: Add america arsenic a preferred source on Google.
ZDNET's cardinal takeaways
- OpenAI partners pinch AMD to utilize GPUs for its AI infrastructure.
- OpenAI CEO Sam Altman says nan AI bubble bursting isn't a concern.
- The AI roar has fueled an unbridled request for computing power.
The fame of generative AI technologies has led to an unprecedented magnitude of investment successful AI-related companies, infrastructure, and solutions. This has sparked concerns astir nan anticipation of an AI bubble bursting and existent ROI connected investments, but OpenAI's CEO, Sam Altman, is now pushing backmost against that idea.
At OpenAI DevDay 2025, Altman told nan property successful a Q&A that while nan AI assemblage has immoderate "bubbly" qualities, which whitethorn lead to overinvestment, nan finance rhythm is simply a normal portion of immoderate technological revolution, and it doesn't contradict nan worth that being built successful AI space.
Also: OpenAI DevDay event: Agent Kit, Apps SDK, ChatGPT, and more
"People will overinvest successful immoderate places," said Altman. "There will beryllium galore bubbles and corrections complete that period, but what I don't deliberation [is that] this is wholly divided from reality -- there's a existent point happening here."
These comments travel astatine a clip erstwhile nan ROI connected AI has yet to beryllium proven, pinch aggregate studies suggesting that contempt hefty investments, nan return remains unclear. For example, an MIT study recovered that 95% of enterprises attempting to harness nan exertion aren't seeing measurable results successful gross aliases growth.
While Gaurav Gupta, Gartner's VP Analyst successful Emerging Trends and Technologies, acknowledges that AI ROI has yet to beryllium proven, he finds that location is simply a agelong roadworthy up until AI solutions, specified arsenic LLMs, scope their afloat potential, which is successful portion fueling further investments.
"We person been talking astir nan AI bubble, wherever enterprises are uncovering it difficult to execute ROI beyond first productivity gains pinch AI," said Gaurav Gupta, Gartner's VP Analyst successful Emerging Trends and Technologies. "On nan different hand, you tin spot hyperscalers, frontier labs, and advertizing companies proceed to walk to get entree to much compute. This tells america that a batch much activity still needs to beryllium done connected LLMs and a title towards AGI -- hence each nan crazy request for compute."
Also: Despite AI-related occupation nonaccomplishment fears, tech hiring holds dependable - and present are nan astir in-demand skills
A premier illustration of a hyperscaler finance was made connected Monday, up of DevDay, erstwhile AMD announced a caller business pinch OpenAI. In this partnership, OpenAI agreed to utilize aggregate generations of AMD Instinct GPUs to powerfulness its AI infrastructure, utilizing six gigawatts of power. The first one-gigawatt deployment of AMD Instinct MI450 GPUs is scheduled to return spot successful nan 2nd half of 2026. As portion of nan deal, AMD granted OpenAI a warrant for up to 160 cardinal shares, balanced to 10% of nan outstanding shares of AMD communal stock.
This woody is only 1 of galore OpenAI has made precocious to meet GPU demand. In September, an Nvidia and OpenAI partnership allowed OpenAI to deploy 10 gigawatts of Nvidia systems, representing millions of GPUs. As portion of nan partnership, Nvidia agreed to put up to $100 cardinal successful OpenAI. Around nan aforesaid time, OpenAI expanded its woody pinch GPU unreality supplier CoreWeave to astir $22.4 billion.
Gupta added that nan willingness of these monolithic companies to make important investments successful OpenAI supports nan premise that OpenAI tin present a batch of worth successful nan AI abstraction successful nan years ahead.
"OpenAI is still a backstage institution pinch monolithic valuation, but large companies for illustration Nvidia and AMD proceed to make deals pinch OpenAI -- it tells a batch astir what they comprehend astir OpenAI -- promising early potential," said Gupta.
The voracious industry-wide appetite for GPUs is fueled by a request for compute, which not only powers existent offerings but besides enables nan improvement of caller ones. Recent merchandise releases person continued to item nan request for much compute power.
Also: AI lifts immoderate package stocks, leaves others down - who's winning and losing and why
For example, Greg Brockman, OpenAI Vice President, said that nan number 1 instruction from nan Sora video generator merchandise is nan request for further compute power. Another illustration is nan caller Pulse feature, which provides users pinch a personalized digest of news from crossed nan web based connected their chat activities pinch ChatGPT. Although it is a highly beneficial feature, its implementation is constricted to Pro subscribers owed to nan computational demands of nan feature. Ultimately, for each further GPU investment, Altman said, location is simply a coming ROI there.
"We tin still monetize each GPU we get our hands connected ace well; I deliberation that will support going for a agelong time," added Altman. "The grade to which we could person 10x of compute, we could build truthful galore much products and connection truthful galore much services group would love."
1 month ago
English (US) ·
Indonesian (ID) ·