- Neoclouds split GPU orders between Nvidia and AMD at rough parity, VAST Data's John Mao said
- AMD's financials support Mao, reporting $5.8 billion in Q1 data center revenue, up 57%, and projected EPYC sales up 70% and Instinct up 46% in Q2
- Mao said the gap between GPU demand and supply widened over the past 12 months, after he expected it to narrow
Nvidia's hold over AI data centers could be slipping. Neoclouds are splitting GPU orders between Nvidia and AMD at rough parity, seeking to take advantage of AMD's price advantage, said John Mao, VP of global technology alliances at VAST Data.
Mao said at least two providers are standing up 10,000 GPUs from each provider side by side.
VAST, which sells the software that feeds data to those GPUs, expanded its AMD partnership last month to sell to buyers of both.
"The performance still swings in the favor of Nvidia, but there's a price advantage to AMD," Mao told Fierce in an interview at the AMD Advancing AI 2026. He attributed that perspective to a neocloud provider he met with.
AMD's results show increased demand. The company reported $5.8 billion in data center revenue for Q1 2026, up 57% year over year, and projected EPYC sales up 70% and Instinct sales up 46% in Q2. Nvidia still has a healthy lead, with roughly three-quarters of AI accelerator revenue projected for 2026, according to Silicon Analysts estimates. However, that's down from a peak near 87% in 2024.
The buyer pool is growing too. Neocloud revenue topped $25 billion in 2025 and Synergy Research Group expects it to near $400 billion by 2031, with McKinsey already counting more than 100 players. That forecast has telcos, hyperscalers and private equity chasing the business.
No matter whose processors VAST's customers buy, they're reckoning with long lead times. "Everybody has constraints right now," Mao said. "Nvidia's got long lead times. AMD's got long lead times. We've got long lead times." He said he previously expected that problem to mitigate over time, but instead it's gotten worse. "The gap has widened in the last 12 months, not converged," he said.
What VAST and AMD agreed to
The July 23 announcement was the first go-to-market deal between the two companies, though VAST has used AMD silicon for years.
VAST selected 6th Gen AMD EPYC processors, formerly code-named Venice, for the sixth generation of its CBox platform and the third of its EBox. AMD said the parts double I/O bandwidth generationally through PCIe Gen 6 and cut latency for VAST's database and event-streaming services.
CPUs matter to AI because agentic workflows need conventional processors for orchestration and tool-calling. AMD's internal projections put the CPU-to-GPU ratio on a path toward 1-to-1, down from earlier estimates near 4-to-1, Mao said.
What's the telco role? CEO Renen Hallak has said operators have been "irrelevant for a very long time" after ceding cloud and streaming to AWS and Netflix, and argues AI hands them a second chance through edge sites and sovereign infrastructure — the same buyers now ordering AMD and Nvidia in equal measure.
VAST raised $1 billion at a $30 billion valuation in April, generates more than $100 million in cash a quarter and expects to be IPO-ready this year without committing to filing.
Learn more about VAST Data
VAST seeks to close the AI trust gap
Enterprises now write $100M checks for AI infrastructure, says VAST Data VP