Rajiv Ramaswami’s grinning like a cat with a Broadcom-shaped canary.
He’s at Nutanix NEXT in Chicago, dangling mid-teens revenue growth by fiscal ‘29 — assuming geopolitics doesn’t implode. Wall Street laps it up. But here’s the kicker: a big slice comes from swiping VMware customers, those 165,000 souls abandoned in Broadcom’s acquisition circus.
Nutanix already nabs 500 to 1,000 quarterly. Add that to their 30,000 base, and it’s a feeding frenzy. “We expect that there’s still a lot of opportunity [with Broadcom], and it’s going to come in waves,” Ramaswami says. Waves. Cute. Like VCF 9 won’t trigger another exodus.
Zoom out. This VMware poach-a-thon funds Nutanix’s real bet: GPU virtualization for enterprises. Yeah, GPUs need it as bad as CPUs did back in the VMware glory days. Remember when server sprawl was the devil? Virtualization tamed it. Now AI inference chews GPUs like candy, but scarcity bites. Enterprises can’t buy ‘em fast enough — or share ‘em smartly.
Why Do Enterprises Crave GPU Virtualization Now?
GPU scarcity. That’s the dirty secret. Nutanix CTO Mano Bhattacharyya spills it: customers run AI on Nvidia, AMD GPUs, even CPUs with tiny models. Why? No GPUs to be had.
“They’re doing it on Nvidia GPUs,” Bhattacharyya said. “They’re doing it on AMD, and lot of them are also doing it on CPUs, specifically with small language models or small transformers, because what is happening is GPU scarcity is a big issue now, specifically in the enterprises.”
Spot on. Frontier models hog hyperscalers; enterprises scrape by with inference basics — document search, fraud hunts. Nutanix drops a multi-tenancy framework at GTC, then amps it this week. Squeeze more juice from fewer GPUs. Share ‘em across tenants, neoclouds, whatever. Sounds VMware-esque, doesn’t it? Irony: Nutanix weaponizes virtualization against its old rival, just as GPUs mirror CPU chaos from 20 years ago.
My unique twist? This echoes Sun Microsystems’ dot-com bust. They ignored x86 virtualization; VMware ate their lunch. Broadcom-VMware? Smells like Sun 2.0 — proprietary walls cracking under AI gravity. Nutanix positions as the open(ish) savior. Bold call: if they nail GPU virt, they own enterprise AI infra by 2027.
But wait. Ramaswami admits customers aren’t there yet. Agentic AI? Years off. Sovereignty, data locale, agent autonomy — mission-critical stuff demands caution. “I think it’s going to be few years,” he shrugs.
Poaching pays the bills.
Is Nutanix’s Agentic AI Platform All Sizzle, No Steak?
Nutanix loves its internal AI glow-up. Engineering? 20% dev productivity boost. “We think we can target something like a 20% productivity improvement per developer,” Ramaswami boasts. Means more features, same headcount. Support teams juggle extra cases sans hires. Bottom-line use, baby.
“What we’re doing is to say, look, that means we can get features out faster without increasing teams.”
Smart. AI in Nutanix hums; AI on Nutanix lags. Customers poke at simple inference. Agentic era? Dozens, not thousands, deploy it meaningfully. Platform strategy unveiled last month — multi-tenancy for GPUs chief among features. Partners with Nvidia (now enterprise-curious for inference) and AMD (who invested). But rollout? Early days.
Corporate hype alert. Ramaswami splits hairs: AI on Nutanix (customer workloads) vs. AI in Nutanix (internal). Latter juices margins; former’s a tease. Enterprises grapple GPU famine, flirt with CPUs. Nutanix pitches private datacenters, flexible models. Sovereignty sells. Yet, “relatively simple inferencing use cases” dominate production. Not agentic. Not yet.
Dry humor: It’s like promising warp drive while customers fix flats with duct tape.
Nutanix isn’t blind. They foresee waves — VMware defections first, then AI maturity. Geopolitics stable? High-20s margins. Capricious Broadcom spooks long-termers eyeing AI. VCF 9? Defection trigger.
GPU virt: the quiet killer app.
But will it scale before hype deflates?
Why Does GPU Virtualization Echo CPU History — And Trap Broadcom?
Flashback. CPUs went virtual; efficiency exploded. Multi-tenancy slashed sprawl, costs. GPUs? Same script. AI inference — production today — guzzles ‘em. Scarcity forces CPU fallbacks, smaller models. Nutanix framework virtualizes: slice, dice, share securely.
Enterprises win: max utilization, no silos. Neoclouds too. Nvidia shifts enterprise-ward; AMD ramps. Nutanix glues it. Critique: PR spin screams ‘me too’ — but timing’s gold. VMware’s Broadcom baggage? Perfect storm.
Prediction: Nutanix hits 50,000 customers by ‘27, half ex-VMware. GPU virt becomes table stakes, like CPU virt post-2005. Broadcom fumbles AI pivot; Nutanix feasts.
Still, agentic AI waits. Customers test waters — fraud, summaries. Production? Basic. Autonomy scares for mission-crit. Few years, sure. But GPU virt? Deploy now, win later.
Ramaswami’s measured. No moonshots. Internal wins fund patience. Investors buy it.
Skeptic’s note: Dozens of meaningful AI users? That’s beta, not boom. VMware poach? Easy pickings, but saturation looms.
Nutanix plays long game smartly — mostly.
🧬 Related Insights
- Read more: Co-Processors: AI’s Secret Weapon in the Chip Wars
- Read more: Optical Metamaterials Swap Cloaks for AI Data Center Speed Demons
Frequently Asked Questions
What is Nutanix’s GPU multi-tenancy framework?
It’s virtualization for GPUs, letting enterprises and neoclouds share scarce hardware securely across tenants — much like CPU virt tamed servers decades ago.
Why is GPU scarcity hitting enterprises hard?
AI inference demand outstrips supply; hyperscalers hoard Nvidia/AMD GPUs, forcing firms to CPUs or small models for tasks like fraud detection.
Will Nutanix steal VMware customers en masse?
Already grabbing 500-1,000 quarterly from Broadcom’s mess; VCF 9 and AI ambitions could wave more over, targeting 165,000 potentials.
