Nexus bets AI's biggest winners will be built beneath the chatbots

The venture capital firm's investments in TensorWave and Neysa reflect its belief that compute, cloud, power and security infrastructure will capture AI's long-term value

Nexus Venture Partners is wagering that artificial intelligence's (AI’s) biggest winners won't be the chatbot makers dominating headlines, but the compute, Cloud and security platforms running underneath them.

 

The venture capital (VC) firm — an early backer of AMD-based Cloud provider TensorWave and Indian graphics processing unit (GPU) startup Neysa — has built much of its recent portfolio around what it calls AI's “infrastructure layer”: The chips, data centres and power systems needed to scale enterprise adoption.

 

“The chatbots are the medium consumers engage with today,” Nexus vice-president Utkarsh Malhotra told Business Standard. “But it's hard to imagine any job, in any sector, that won't be meaningfully altered by AI within a decade.”

 

Nexus is evaluating investments across four areas — vertical AI applications in financial services, healthcare, insurance and legal services.

 

Across all four, the firm said it is looking for global category leaders.

 

That ambition is backed by capital. Nexus manages $3.2 billion across its funds and recently closed a $700 million Fund VIII, which will back early-stage startups in AI, enterprise software, consumer and fintech across both India and the US through a single team.

 

The firm argues the AI cycle is moving faster than prior technology waves such as Cloud computing, which has sharpened geographic advantages: San Francisco currently holds a disproportionate share of frontier AI research, capital and company creation.

 

But Nexus expects AI to be a decades-long transformation rather than a short-lived trend, and believes India will play an increasingly important role as the country applies AI across its economy and produces founders who build globally-significant companies.

 

Nexus's recent AI bets span the stack — infrastructure plays TensorWave and Neysa; application companies Giga, Avoca and Aurva; developer platforms Firecrawl, Bolt.new and Gumloop; and security company Singulr.

 

These reflect a strategy of backing AI-native founders across compute, tooling and security in both the US and India.

 

The infrastructure thesis dates back to the earliest days of the generative AI boom, when Nexus concluded that compute demand would surge and customers would want alternatives to Nvidia — and that AMD's public roadmap and memory architecture suited emerging AI workloads.

 

That conviction led the firm to lead TensorWave's seed round and co-lead Neysa's seed funding, betting India would emerge as a global infrastructure hub.

 

“The world was quite hungry for an Nvidia alternative,” Malhotra said. “The beauty about some of these larger chip companies is that their roadmap is quite public for the next few years. So, we could already see that some of AMD's architectural decisions were favourable for the upcoming AI workloads.”

 

Neysa's own trajectory illustrates the bet. The company went from a $128 million valuation to a $1.4 billion unicorn in about 16 months, with Blackstone and co-investors agreeing in February to a $1.2 billion capital raise. This will fund a planned deployment of more than 20,000 GPUs in India.

 

That scale-up sits against a backdrop in which India reportedly has fewer than 60,000 GPUs deployed today, against a projected long-term need for more than 2 million.

 

This raises the question of whether the country's buildout is a genuine sovereign-capability story or largely a re-export of Nvidia or AMD hardware.

 

Nexus argues it's the former. Operating large compute clusters is a complex technical challenge requiring specialised expertise, capital and power infrastructure, not just GPU access.

 

Demand remains concentrated among frontier labs and hyperscalers, the firm said, and while India trails today, its talent and growth trajectory position it to become a meaningful hub over time.

 

“Setting up GPUs is way more complex in a new Cloud than previously,” Malhotra said. “These things are fragile, making them work together is a complex technology endeavour, and there are not too many experts in the world.”

 

The economics of AI, he said, have shifted from training to inference as AI agents take on real enterprise work, lifting demand for real-time compute.

 

Industry bottlenecks have moved accordingly — from GPU shortages a year ago, to memory and CPUs, and now to electricity and power.

 

Nexus expects the next phase of infrastructure innovation to focus on expanding energy supply and alternative data-centre designs.

 

“Ultimately, we are going to just generate as much intelligence as energy available for it,” Malhotra said.

 

Asked how founders should build a defensible moat in a market now drawing capital from Blackstone, AMD Ventures and sovereign wealth funds, beyond simply securing GPUs first, Malhotra said differentiation lies in operating large-scale clusters reliably for demanding customers. Only a handful of teams globally have that expertise, he said, and founders should build moats by solving for a small set of tier-one AI labs and hyperscalers rather than relying on early hardware access.

 

“AI’s interface may be a chatbot today, but some of its most enduring value will be built in the infrastructure underneath it,” said Malhotra.