Nvidia Rival Graphcore Makes Big North American Partner Push: Exclusive

The AI chip startup is taking aim at Nvidia’s data center GPU business with its IPUs and purpose-built AI systems, so it’s recruiting solution provider powerhouse Trace3 and other channel partners to ramp up sales in North America. ‘We’re growing both the company and the partner network at a really rapid pace, and that’s due to the fact that the market is absolutely hungry for options that are different than GPUs,’ says Graphcore’s channel chief.

AI chip startup Graphcore is making a big North American channel push this year with new and existing partners to ramp up competition against Nvidia’s data center GPU business.

The Bristol, U.K.-based company exclusively told CRN that it has added three new North American partners — solution provider powerhouse Trace3 and high-performance computing system integrator Applied Data Systems in the United States and solution provider Images et Technologie in Canada. At the same time, three channel partners based in Europe — 2CRSi, BIOS IT and Boston Limited — are expanding their Graphcore reselling efforts to North America.

[Related: 10 Hottest Semiconductor Companies To Watch In 2021]

The roster and territory expansion of Graphcore’s Elite Partner Program, which launched last fall, comes as the chipmaker is making a big sales push this year for its AI-focused Colossus MK2 intelligence processing units as well as the purpose-built M2000 systems and IPU-POD clusters that use the specialty chips. The company’s go-to-market and research and development efforts received a boost late last year when it raised a $222 million funding round that brought its valuation to $2.77 billion.

The new partners — which includes Trace3, ranked No. 39 on CRN’s Solution Provider 500 list — reflects the important of partners to Graphcore’s business, an executive told CRN, and it builds upon existing partnerships in North America with solution providers Meadowgate Technologies and Wildflower International, cloud services provider Cirrascale Cloud Services, Dell Technologies’ OEM Solutions Group and IT distributor Tech Data.

“The partner program is a critical part of our go-to-market strategy, and having the expertise of the partners, their understanding of our technology and their ability to bring our technology into existing customers but also help innovators in the AI space is absolutely a key part of our strategy,” said Victoria Rege, Graphcore’s director of strategic partnerships and alliances, in an interview with CRN.

The benefits of Graphcore’s Elite Partner Program include deal registration, which gives partners margin on the company’s products, as well as resources for sales enablement, training, marketing and support. The company plans to eventually add sales and technical certifications to its program that will cover the company’s IPU chips and hardware as well as its Poplar software development kit.

“That’s the next area that we’re looking at deploying and rolling out for them, which, to me, is a great sign — the fact that the sales leaders and the technical engineers are saying, ‘hey, how do I get certified? How do I get official officially certified on IPU or certified to support the software?” said Rege, who previously worked in ecosystem, marketing and sales roles for nine years at Nvidia.

Graphcore has made it clear in past announcements that it’s going after Nvidia’s fast-growing data center GPU business with claims that its IPU architecture is better suited for AI applications because it was built from the ground up for “fine-grained parallelism” while also coming with 900 MB of on-board ultra-high-speed RAM, allowing the IPU to hold large AI models inside the memory. The company has also claimed that its M2000 systems provide more performance-per-dollar than Nvidia’s DGX systems.

Rege said Graphcore has found that its IPUs can outshine GPUs for AI workloads like natural language processing, computer vision, recommenders and probabilistic learning. She added that the company has its own research team, which keeps tracks of emerging use cases that can benefit from IPUs.

“We’re growing both the company and the partner network at a really rapid pace, and that’s due to the fact that the market is absolutely hungry for options that are different than GPUs,” Rege said.

Several of Graphcore’s North American partners said they see a lot of promise with the company’s IPUs in the AI space. That includes Matt Fornito, head of AI at Campbell, Calif.-based Trace3.

“At Trace3 our customers are looking for us to solve business problems by leveraging deep learning to harness their data. One of the technology tools we bring to the table is Graphcore. We find its ability to find new efficiencies in today’s most demanding workloads is a terrific fit in our AI solutions,” said Fornito, whose company is also an Nvidia partner.

One of Graphcore’s earliest partners is Cirrascale Cloud Services, whose CTO, David Driggers, said he expects 2021 will be the chipmaker’s first big sales year where customers will be more focused on buying systems for development and production versus testing.

“Here we’re looking at some real development and ideally, before the end of the year, some real production or at least pre-production,” he told CRN.

Driggers said Graphcore’s IPUs are better suited than GPUs to handle HPC workloads like risk analysis and genomics that need to analyze a high volume of data points, otherwise known as sparse data. That’s because of the IPU’s “fine-grained parallelism” and its high-speed memory access, he added, which could make the new kind of computer chip a game-changer.

“That type of modeling where you’re trying to deal with what are billions or trillions of parameters and looking for patterns, that’s an area where GPUs have never caught on because of how you interact with the data,” he said.

These specialized capabilities will allow Graphcore to carve out its own niche in the AI space, according to Driggers. But that doesn’t mean the market opportunity for the chipmaker is small, he said. Instead, he expects the demand for AI applications to continue to grow dramatically in the years to come, creating the need for different kinds of accelerators like Graphcore’s IPUs to service different segments.

“When I say niche, it’s 10 percent [of the market]. That, to me, reflects a niche,” Driggers said. “But if you look at how big the opportunity is, 10 percent is still a [really] big number.”