The IT industry continues to adapt to the insatiable demand of generative AI workloads for compute power.
Samsung Electronics executives this week said they will push the chip foundry, memory, and advanced packaging technology (AVP) teams to more quickly push silicon offerings out into a market that is looking for more alternatives to the highly popular, but hard-to-get, GPUs from Nvidia, which account for as much as 75% to 90% of the AI processor market.
At the same time, OpenAI and Microsoft this week announced a deal with Oracle that will allow the ChatGPT maker to use Microsoft’s Azure AI platform on Oracle’s cloud infrastructure, giving OpenAI CEO Sam Altman the compute capabilities he said his company needs to power its generative AI services.
Using the database giant’s Oracle Cloud Infrastructure (OCI) “will extend Azure’s platform and enable OpenAI to continue to scale,” Altman said in a statement.
The Need for Compute
Both moves come as generative AI training and inferencing workloads are expected to scale the demand for compute power in the coming years and as major cloud hyperscalers like Amazon Web Services (AWS), Google and Microsoft develop their own AI processors to ensure they have their own supplies. Established chip players like Intel, AMD, Arm and Qualcomm are pushing out AI chips and smaller startups are looking to gain footing as alternatives to Nvidia.
Samsung’s announcement was one of a number made during its Samsung Foundry Forum (SFF) U.S. at its Device Solutions America headquarters in San Jose, California. Executives said its new turnkey Samsung AI Solutions platform, which brings its foundry, memory chip and AVP services together, will allow it to offer products that offer high performance and bandwidth and low power that can be customized to meet the specific AI needs of customers.
In addition, such collaboration will also streamline the vendor’s supply chain management and reduce the time to market, enabling Samsung to get its AI chips into the market 20% faster.
“We are truly living in the age of AI,” Siyoung Choi, president and general manager of Samsung’s foundry business, said during the event, according to Reuters. “The emergence of generative AI is completely changing the technology landscape.”
Choi said Samsung expects the AI processor demand to drive revenues in the global chip industry to $778 billion by 2028.
Fast-Growing AI Chip Market
The global AI chip market itself will see explosive growth in the coming years. According to market research firm Statista, the space, which was $23 billion last year, will grow another 30% this year before surpassing $150 billion by 2030.
Statista analysts noted the need for specialized and powerful AI chips for everything from image recognition and natural language processing to recommendation engines. They also will make their way into consumer devices, as illustrated by the focus on AI PCs in recent months and Apple’s partnership with OpenAI announced this week that will see AI technologies integrated into its devices and operating systems.
Also at Samsung’s event, company executives also announced two new process nodes designed for AI chips, the 2-nanometer SF2Z set for full production in 2027 and the 4nm SF4U, which will go into production next year.
“At a time when numerous technologies are evolving around AI, the key to its implementation lies in high-performance, low-power semiconductors,” Choi said in a statement, adding that the company will “introduce integrated, co-packaged optics (CPO) technology for high-speed, low-power data processing, providing our customers with the one-stop AI solutions they need to thrive in this transformative era.”
That is expected to roll out in 2027.
OpenAI’s Need to Scale
For OpenAI, the announcement with Oracle is an important step to allow it to continue to scale its portfolio. Altman has been outspoken about the need for more compute power, going so far as to propose the company spend billions of dollars to build its own chip fabrication facilities.
Microsoft, with its Azure AI platform, has been OpenAI’s only cloud partner. Microsoft has invested $13 billion in the company and owns just less than half of it and has integrated the smaller vendor’s technologies throughout its product portfolio.
In a post on X (formerly Twitter), OpenAI said its relationship with Microsoft hasn’t changed with the agreement with Oracle.
“The partnership with OCI enables OpenAI to use the Azure AI platform on OCI infrastructure for inference and other needs,” the company wrote. “All pre-training of our frontier models continues to happen on supercomputers built in partnership with Microsoft.”
Oracle and AI
Oracle Chairman and CTO Larry Ellison said in a statement that “the race to build the world’s greatest large language model is on, and it is fueling unlimited demand for Oracle’s Gen2 AI infrastructure. Leaders like OpenAI are choosing OCI because it is the world’s fastest and most cost-effective AI infrastructure.”
Oracle said the company is joining “thousands of AI innovators across industries worldwide” to run their workloads on the OCI AI infrastructure, listing such vendors as Nvidia, Together AI, Twelve Labs, and xAI, Elon Musk’s 11-month-old generative AI company that recently got a $6 billion boost during a Series B funding round.
Musk, a founder and early investor of OpenAI until leaving the company in 2018, this week asked a California state court in San Francisco to dismiss the lawsuit he filed against OpenAI in February. Musk accused Altman and other company founders of violating a founding agreement to keep the company a nonprofit by partnering with Microsoft and takings its billions of dollars.
He didn’t give a reason for dismissing the suit, though he wants to do so without prejudice, which would allow him to refile it in the future.