About two decades ago, Nathan Brookwood was advising giant chip maker Intel, which the Federal Trade Commission was looking into for allegedly abusing its monopoly power in the fast-growing processor market. Intel executives’ argument at the time was that despite the vendor’s 80-plus percent share of the PC chip space, it was a highly competitive market and that if the company didn’t work hard, it would easily lose market share.
“That was a tenuous [argument],” Brookwood, principal analyst for Insight 64, told InternetNews. “Fast forward 20 years and it’s a very competitive market.”
The last couple of weeks have again highlighted how competitive and volatile the processor market is these days. The CEOs of Nvidia and Arm pushed back at criticism of Nvidia’s planned $40 billion acquisition of Arm, saying it would be difficult for Arm to keep up with increased demand on its own. Meanwhile, the incoming president and CEO of Qualcomm – a vocal opponent of the deal – suggested that if the controversial sale doesn’t go through, Qualcomm and other vendors who make Arm-based processors could invest in Arm, giving it a surge of cash.
The Qualcomm executive, Cristiano Amon, and Intel CEO Pat Gelsinger, both were panelists at a CNBC conference and talked about the possibility of the two rivals collaborating, noting the former’s expertise in communications technology and the latter’s strength in computing.
Competitors on the Move
In the meantime, other Intel competitors continue to make strides. Google Cloud announced the launch of a new family of virtual machines – dubbed Tau – that will be built on AMD’s third-generation “Milan” Epyc processors. And Oracle Cloud said it now is offering instances powered by Ampere’s Arm-based Altra Max chips and launching its Arm Accelerator program to support application development for Arm processors.
There are a number of drivers fueling the rapid changes and consolidation in the data center chip space, particularly accelerating shift to the cloud, which Brookwood said “has really upended a lot of the assumptions and that drove the market before.” Enterprises traditionally have bought servers and installed and managed them in their data centers to run third-party software. The cloud is changing that.
“As [businesses] move to the cloud, the economics change because [Intel] can’t charge a ton for the processors that go with the cloud servers compared with what they could charge for the processors that went into enterprise servers,” he said. “Also, if you look at companies like Amazon or Oracle, they have the ability – since they control most of their own software stack – to change the underlying processor architectures whenever it makes sense for them, either from an economic standpoint or performance standpoint.”
That combined with other factors – from manufacturing missteps by Intel and the slowing down of Moore’s Law to the desire of enterprises for more processor options in the servers they buy – have created an environment where Intel needs to compete to keep market share.
Further reading: IBM’s 2nm Chip Will Remake IT Industry
Intel Pushes Back
The company is trying to do just that. Earlier this year it brought in Gelsinger – who had decades of experience at Intel before becoming the CEO of VMware in 2012 – as CEO. He has since overseen Intel’s bold move to double down on its manufacturing capabilities, including investing $20 billion on new fabrication facilities. More recently, the CEO last week announced a reorganization inside the company that includes creating two divisions within the Data Platform Group – one for data centers and artificial intelligence (AI) and the other for networking and the edge – and putting new executives in charge.
“You don’t go and make major changes like that when everything’s going hunky-dory, so clearly Gelsinger came in, he assessed the situation and he said, ‘Boy, we have some problems here’ and is shaking up the organization,” Brookwood said.
At the same time Gelsinger is reaching out, reports began circulating earlier this month that Intel is considering spending $2 billion to buy SiFive, a startup that is building chips based on the open-source RISC-V architecture and is looking to compete with Arm. Intel officials declined to comment on the reports. However, the two companies this week announced an alliance that will see Intel manufacture SiFive’s new P550 core in its 7-nanometer process. Intel will put the SiFive cores into a chipset codenamed “Horse Creek,” which chip makers will use to evaluate whether to use the SiFive cores in products.
Gelsinger’s comments with Qualcomm also show his openness to working with other manufacturers.
“We are the unquestioned compute leader and Qualcomm’s the unquestioned comms leader,” he said at the CBNC event. “Compute meets comms. A lot of new use cases.”
AMD, Arm and the Cloud
Those use cases could be important as Intel navigates a market that includes a reinvigorated and competitive AMD and Arm, which not only could see an infusion of money if Nvidia’s $40 billion is allowed but also is being embraced by cloud providers. Oracle Cloud’s announcement that it will offer instances powered by Ampere’s 128-core chip is a significant step forward for Arm’s long-held data center ambitions.
“We’re seeing interest from early customers that are using our new Oracle Cloud Infrastructure Ampere AI Compute instances for computationally intensive workloads like machine learning inferencing, encoding and HPC [high-performance computing],” Matt Leonard, vice president of compute at Oracle Cloud Infrastructure, said in a statement.
Insight 64’s Brookwood said “the whole thing with Ampere and Arm at Oracle, that’s kind of amazing in the sense that we’ve been talking about Arm in the data center for practically a decade and finally, you could point to one sort of emerging success. I wouldn’t call it a done deal yet, but it’s emerging. Oracle’s no slouch.”
Both Qualcomm and Intel oppose Nvidia’s proposal to buy Arm from Japanese multinational Softbank. Amon has argued that given that broad range of manufacturers that rely on Arm architectural designs – which it licenses to chip makers, including Nvidia and Qualcomm – it’s important that Arm remains independent. He has suggested that Qualcomm and other industry investors would take a stake in Arm if the Nvidia deal falls through.
Nvidia, Arm CEOs Make Case for the Deal
However, speaking with Patrick Moorhead, principal analyst at Moor Insights and Strategy, at the recent Six Five Summit, Nvidia CEO Jensen Huang and Arm CEO Simon Segars said Arm would need more than that to address the rapid increase in demand for computing.
“The range of products that our licensees want to build is growing and growing,” Segars said, noting not only the money but also the resources Nvidia brings to the table. “What they’re asking from us is increasing and increasing because of the complexity going up. There’s no way that we could do it on our own.”
The Need for Federal Help
All of this comes amid an ongoing push by the Biden Administration and Congress to ramp up processor-making capabilities in the United States to help alleviate an ongoing worldwide shortage of chips and to make more processors in this country. Most processors are now made in Asia by either the Taiwan Semiconductor Manufacturing Corp. (TSMC) or Samsung.
According to the Semiconductor Industry Association (SIA), the United States’ chip manufacturing capacity has dropped from 37 percent in 1990 to 12 percent this year. The industry trade group said a combination of investments by other governments in manufacturers in their countries and inadequate investment by the United States has helped lead to the imbalance.
However, lawmakers are hoping to change that. In a rare bipartisan effort, the Senate earlier this month passed the CHIPS (Creating Helpful Incentives to Product Semiconductors) Act that included $52 billion in federal investments for chip research, design and manufacturing in the United States.
President Biden and U.S. lawmakers have positioned the effort as an important step for improving the United States’ technology competitiveness against China and for mitigating the worldwide chip shortage that has impacted not only tech system manufacturers but other sectors like the automotive industry, where car makers have been unable to buy enough semiconductors, forcing them to cut back on the number of cars they’re making.