The race to build specialized chips for running trained AI models is heating up, with startups raising billions in funding and attracting major investors. Positron, an energy-efficient inference chip startup, is in talks to raise approximately $750 million in a two-phase funding round that could value the company at $5 billion within months. This surge reflects growing investor appetite for companies focused on inference, the stage where AI models respond to real-world user inputs after training is complete.

Why Are Investors Suddenly Betting Billions on Inference Chips?

The inference chip market has become a focal point for venture capital and strategic investors because it represents a distinct opportunity from the training chips that Nvidia currently dominates. While training requires massive computational power to build AI models from scratch, inference is the ongoing process of using those trained models to answer questions, generate text, or process data. This creates a different set of technical requirements and a potentially enormous market as AI applications scale globally.

The funding momentum is undeniable. SambaNova Systems announced a $1 billion funding round at an $11 billion valuation this week, while Etched, another Nvidia competitor, recently raised $800 million and signed sales contracts worth $1 billion. Cerebras Systems, which also develops inference chips, completed the year's largest initial public offering in May, though its stock has since faced challenges, trading below its IPO price after reaching a high of $311.07.

How Are These Startups Positioning Themselves Against Nvidia?

The competitive landscape is shifting in unexpected ways. Nvidia itself is taking a defensive posture by partnering with competitors rather than simply crushing them. In December, Nvidia agreed to license inference chip technology from Groq and is integrating that technology into its own products. This move suggests that even the market leader recognizes the value of specialized inference solutions and the risk of being outflanked by focused competitors.

Startups are attracting heavyweight investors who see long-term value in the inference market. Positron's investor list includes Valor Equity Partners, Atreides Management, DFJ Growth, Arm Holdings, and the Qatar Investment Authority. These are not typical venture capital firms; they represent strategic capital from semiconductor companies and sovereign wealth funds betting on the future of AI infrastructure.

What Makes Inference Chips Different From Training Chips?

• Energy Efficiency: Inference chips are designed to run trained models with lower power consumption than general-purpose processors, making them ideal for data centers operating at scale.
• Latency Requirements: Inference demands fast response times for real-world applications, whereas training can tolerate longer processing windows.
• Cost Structure: Inference happens continuously across millions of user interactions, making per-operation costs critical to profitability.
• Market Size: As AI applications proliferate, the inference market is expected to grow larger than the training market over time.

What Do Analysts Say About the Inference Chip Market?

Wall Street is taking notice of the opportunity. One analyst rates Cerebras Systems a Buy with a $239 price target, representing 32% upside from the current level of $182 per share. The analyst's thesis centers on the market still deciding what Cerebras really is, with growth driven by OpenAI deployment, AWS distribution, hardware expansion, and margin normalization.

"The core of my thesis is that the market is still trying to decide what Cerebras really is," noted the analyst in their investment thesis.

The Curious Analyst, Seeking Alpha

The analyst's forward earnings estimate of $3.19 per share for 2028 is built on multiple growth drivers, with OpenAI and AWS forming the largest pieces of the earnings bridge. This suggests that major cloud providers and AI labs are actively evaluating and potentially deploying inference chips from companies other than Nvidia.

Why Is This Moment Different From Previous Chip Startup Waves?

Previous attempts to challenge Nvidia in AI chips often failed because they tried to compete across the entire spectrum of AI workloads. Today's startups are taking a different approach by focusing narrowly on inference, where the technical requirements are more specialized and the addressable market is enormous. The funding environment also reflects genuine demand signals from cloud providers and AI companies actively testing alternatives.

The fact that funding rounds are happening in multiple phases in rapid succession underscores the ultracompetitive pressure in the current technology landscape. Positron's two-phase round, with valuations jumping from $3.5 billion to potentially $5 billion, shows how quickly investor sentiment can shift when a company demonstrates traction or secures major customer commitments.

The inference chip market represents a genuine inflection point in AI infrastructure. Unlike previous cycles where one company could dominate, the sheer scale of inference workloads and the diversity of use cases suggest that multiple specialized chip makers will coexist. Startups are raising record amounts because investors believe the inference market is large enough to support multiple winners, and Nvidia's licensing strategy suggests the company agrees.