Cerebras $2.5B: physical AI enters B2B manufacturing

The timeline: from lone bet to a $2.5 billion exit

In 2015, Lior Susan founded Eclipse Ventures with a contrarian premise. While American venture capital chased SaaS and digital marketplaces, Susan bet on the physical world. Robotics, semiconductors, industrial automation: sectors seen as slow, capital-intensive and unglamorous.

Cerebras Systems was one of those investments. The company developed AI chips of unusual size, the Wafer-Scale Engine, designed to speed up training and inference of large models. Its positioning was also explicitly an alternative to Nvidia: fewer commodity GPUs, more architecture specialized for intensive AI workloads.

In May 2026, TechCrunch reported that Eclipse sees this exit as only the beginning. The thesis on the physical world is far from exhausted. On the contrary, it's proving more relevant than ever.

Anatomy of the bet: why silicon matters more than software

Cerebras has built a real competitive advantage. Its chip packs billions of transistors onto a single wafer, eliminating the communication latency between separate chips. For applications that need fast inference, such as in-line quality control, predictive maintenance and computer vision, this approach offers measurable benefits.

The difference from traditional cloud AI matters. Models hosted on remote infrastructure introduce latency, dependence on connectivity and data sovereignty concerns. An on-premises or edge architecture, by contrast, allows local processing, which is essential in industrial settings where production data is sensitive.

So Cerebras's value isn't just financial. It represents the maturing of an entire category: specialized AI hardware for physical applications. According to Gartner, by 2027 more than 40% of enterprise AI workloads will run at the edge or on-premises. The Cerebras deal anticipates that trajectory.

Winners and losers: who really gains from this scenario

The most obvious winner is Eclipse Ventures. The firm has shown that investing in the physical world isn't a romantic niche, but a thesis with concrete returns. The credibility it has gained will also draw capital toward other portfolio companies in AI hardware and robotics.

Cerebras itself strengthens its position as a credible alternative to Nvidia. The road is still uphill, though: according to industry estimates, Nvidia controls over 70% of the AI GPU market. Cerebras's specialization is an advantage in specific segments, not a sweeping market conquest.

The ones at risk of losing ground are SMEs that put off decisions on AI infrastructure. While big industrial players in automotive, aerospace and pharma are already evaluating on-premises AI architectures, mid-sized companies risk falling structurally behind. As a result, the competitive gap could widen over the next 18-24 months.

A Milan agency's view: what we see in the Italian market

At SHM Studio, we work every day with Italian SMEs in B2B and retail. We see a clear trend: demand for AI solutions is growing, but it often runs into a still superficial understanding of the infrastructure required.

Many companies associate AI exclusively with cloud tools like ChatGPT, Copilot and SaaS platforms. But for those in manufacturing, logistics or supply chain, the real opportunity is AI applied to physical processes. That calls for thinking about hardware, not just software.

The Cerebras-Eclipse deal is a market signal that Italian SMEs should read too. Anyone working in digital marketing and digital transformation can't ignore the infrastructure side of AI. The two, digital communication and technology infrastructure, are increasingly intertwined.

Operational implications for Italian manufacturing SMEs

What does this scenario mean in practice for an Italian manufacturer with 50-500 employees? First, it helps to distinguish between three levels of AI maturity.

• Basic level: using cloud AI tools for marketing, copywriting and data analysis. Accessible today, with modest investment.
• Intermediate level: integrating AI models into specific business processes: predictive CRM, customer segmentation, optimization of Google Ads and LinkedIn campaigns. Requires structured digital skills.
• Advanced level: on-premises AI for quality control, predictive maintenance and computer vision. Requires hardware investment and specific skills.

It's also important to assess where you stand today before jumping to advanced solutions. A company that hasn't yet optimized its SEO positioning or its digital presence isn't ready for on-premises AI. Sequence matters.

That's why our approach at SHM Studio always starts with an assessment of overall digital maturity. Only then can you identify where AI creates real value, and where it risks being a premature investment.

The open question: where the game will be played over the next 24 months

Eclipse Ventures' thesis doesn't end with Cerebras. According to founder Susan, the portfolio includes dozens of companies working at the intersection of AI and the physical world. So in the coming years we'll likely see more significant exits in robotics, energy tech and industrial automation.

At the macro level, McKinsey estimates that generative and applied AI could add up to $4.4 trillion a year to the global economy. A significant share of that value will come from industrial and physical applications, not just digital ones.

Italy's context is specific. Its productive fabric is made up mostly of manufacturing SMEs with strong sector specialization. So adopting physical AI, even in forms less sophisticated than Cerebras chips, is a real competitive opportunity. Especially for companies that export and compete in international markets.

Likewise, those working in web development and digital presence need to start thinking about how on-premises AI will change data flows and integrations between physical systems and digital platforms. The line between IT and OT (Operational Technology) is thinning fast.

Next moves: three priorities for those who want to get ahead

Based on this analysis, we can identify three priority areas for action for Italian SMEs that want to prepare for this scenario.

First priority: a digital maturity audit. Before considering any AI investment, you need a clear picture of where you stand today. That includes SEO, web infrastructure, the quality of your internal data, and analytics capabilities. Without that foundation, any AI investment risks being ineffective.

Second priority: in-house training on applied AI. You don't need to become an expert in chips or hardware architectures. What matters is understanding which business processes can benefit from AI, and which skills you need to evaluate vendors. Training also lowers the risk of decisions driven by hype instead of analysis.

Third priority: choosing your technology partners. The B2B AI market is crowded with solutions. Choosing partners with vertical experience in your industry is essential. Finally, it pays to check whether the proposed solutions are scalable, meaning they can grow with your company without requiring costly migrations.

To dig deeper into these topics or start an assessment of your digital position, you can contact the SHM Studio team or explore the content on our blog.