Private Market Company Spotlight: Cerebras Systems — The Apex Play of the AI Infrastructure Era
“We Don’t Chase Trends, We Identify Generational Companies Before The World Catches On. Cerebras Systems Is An Example of a Company that is Solving A Major Problem For their Customers, We Believe Their IPO Will Be A Massive Success, And Has A High Probability Of Producing Exponential Growth Longterm.
— G.L.A. INVE$T RESEARCH
The Macro Signal You Cannot Ignore
The AI hardware landscape just experienced a tectonic shift. The landmark ~$20 billion licensing agreement between NVIDIA and inference specialist Groq — in which NVIDIA essentially paid a fortune to license Groq’s inference technology and absorbed its top leadership — is not just a business transaction. It is a strategic confession. The world’s most dominant chip company just told the market: GPU architecture alone cannot win the inference era.
Inference — the process of running a trained AI model in real time — is rapidly becoming the most economically significant workload in all of computing. In 2025, inference represented approximately 50% of AI computation; by 2026, it is projected to account for 66% of all AI compute. The companies architected from day one to dominate this workload are not NVIDIA imitators. They are a different breed entirely. And Cerebras Systems sits at the top of that list.
🔥 G.L.A. TOP 10 PRIVATE COMPANIES — CEREBRAS RANKED #3
Following the SpaceX / xAI mega-merger — a $1.25 trillion consolidation that repositioned the private company landscape entirely — and Cerebras‘ own explosive $23 billion Series H valuation, our team has formally updated the G.L.A. INVE$T RESEARCH Top 10 Favorite Private Companies list.
Cerebras Systems now holds the #3 position.
This is not a reactionary ranking. It reflects years of tracking this company’s technological moat, its revenue trajectory, its customer depth, and the macro forces now converging in its favor. We will be transparent: we are extremely excited about the upcoming Cerebras IPO, targeted for as early as April–Q2 2026. But the IPO is almost secondary to the larger thesis. The public listing is simply the moment the broader market gets to participate in what smart money has already been pricing in for years.
The Company: Built From a Clean Sheet of Paper
Cerebras was founded in 2016 by CEO Andrew Feldman and a team of veteran chip architects. Feldman, who previously built and sold data center company SeaMicro to AMD, had a singular insight after his exit: he was bored — and more importantly, he saw something the rest of the industry was missing.
“We saw AI on the horizon… And we knew that the graphics processing unit would probably not be the right machine. If we started with a clean sheet of paper and designed a solution optimized for AI — not for graphics, not for databases, not for web serving, but just for AI — we could build a better solution.”
— Andrew Feldman, Co-Founder & CEO
The conviction was validated immediately. When Feldman and his team went to raise their first round in March 2016, they made eight presentations and received eight term sheets. That is not a typical fundraise. That is a founding team and an idea so compelling that the venture community responded in near-unanimity.
The result of that conviction is the Wafer-Scale Engine (WSE-3) — the world’s largest single chip, packing 4 trillion transistors, delivering up to 125 petaFLOPS, and boasting approximately 20x the memory bandwidth of an NVIDIA H100. It was engineered for one purpose: to make AI inference and training dramatically faster and more efficient than anything GPU-based systems can offer.
The Investors Who Saw It First
When the smartest money in the world clusters around a single private company across multiple funding rounds, that pattern is a signal. Here is who has been backing Cerebras — and doing so aggressively:
The Series H round — closed in February 2026 at a $23 billion valuation (nearly triple the $8.1B valuation from just months prior in September 2025) — was led by Tiger Global with participation from Benchmark, Fidelity Management & Research Company, Atreides Management, Alpha Wave Global, Altimeter, AMD, Coatue, and 1789 Capital, among others. Benchmark Capital has reportedly also raised $225 million in special funds specifically to double down on Cerebras — a rare and exceptionally bullish signal from one of Silicon Valley’s most disciplined firms.
This is not tourist capital. Tiger Global, Altimeter, AMD, Fidelity,1789 Capital are strategic, high-conviction institutions that do deep diligence. When they all congregate in the same round at an accelerating valuation, that is the private markets equivalent of a unanimous buy signal.
Fundamentals For the weighing Machine (Longterm Markets)
Cerebras is not pre-revenue speculation. It is a business with real, large-scale enterprise traction:
G42 Partnership (UAE): Beginning with a demonstration in 2023 that solved in days what GPUs had taken months to compute, the partnership exploded. Cerebras built hundreds of exaFLOPS of computing infrastructure for G42, trained the leading Arabic-language model Jais with MBZUAI, and G42 consumed nearly everything Cerebras could manufacture in 2023 and 2024. This single partner represented 87% of Cerebras’ revenue in H1 2024. The Condor Galaxy supercomputer, launched jointly in 2023, became one of the world’s most powerful AI training systems.
OpenAI Partnership: Cerebras recently secured a landmark deal with OpenAI — reported to be worth over $100 billion — to provide 750 megawatts of computing power through 2028. This is the clearest possible signal that even OpenAI, the company most closely tied to NVIDIA’s ecosystem, is actively diversifying its hardware stack.
2025 Enterprise Expansion: Meta, IBM, and Mistral joined as major customers in 2025, marking a decisive broadening beyond a single anchor relationship.
Why This Is an APEX Long-Term Investment
At G.L.A. INVE$T RESEARCH, we use the term APEX sparingly. It designates a company we believe has the architecture — technological, financial, and strategic — to potentially reach trillion-dollar status. Cerebras earns that designation for the following reasons:
Architectural moat: The WSE-3 is not incrementally better than GPU alternatives — it operates on a fundamentally different design philosophy that cannot be easily replicated by retooling existing GPU production lines
Inference tailwind: As AI shifts from training to inference-dominated workloads (66% of all AI compute by end of 2026), Cerebras’ architecture becomes more — not less — advantageous over time
Sovereign AI demand: G42’s use case previews a massive global market — nations and large enterprises building their own AI infrastructure on non-NVIDIA hardware for geopolitical, cost, and performance reasons
Valuation trajectory: $8.1B → $23B in under five months represents a company that the market is actively repricing toward public market comparables
IPO catalyst: A public listing in Q2 2026 will provide liquidity, increased visibility, and a new institutional buyer base that has been locked out of the private rounds
Competitive moat reinforced by NVIDIA’s own moves: When NVIDIA pays ~$20B to license your competitor’s inference technology, the entire inference hardware category gets permanently validated
The IPO: What to Watch
Cerebras has confidentially filed for a U.S. IPO targeting a public listing as early as April 2026. An earlier IPO attempt in 2025 was withdrawn, but the regulatory overhang — primarily related to the G42 ownership structure — has since been resolved. The company enters this IPO window with $1 billion in fresh capital, one of the most recognizable enterprise customer rosters in AI hardware, and a valuation narrative that is accelerating rather than decelerating.
We will be publishing our full IPO Preview Report ahead of the listing with price target frameworks, comps analysis, and a recommended positioning strategy for G.L.A. members.
Final Thoughts…
The GPU era is not ending — but its monopoly on AI compute is. The NVIDIA-Groq licensing deal confirmed it. The OpenAI-Cerebras $100B+ partnership confirmed it. The xAI-SpaceX merger creating an unprecedented $1.25 trillion entity confirmed that the AI infrastructure arms race is entering a phase where scale, speed, and architectural efficiency will determine the winners.
Cerebras Systems — built on a clean sheet of paper, backed by the sharpest capital in the world, and now approaching the public markets — is positioned to be one of the defining companies of this decade.
Ranked #3 on the G.L.A. Top 10 Private Companies. APEX designation. Long-term conviction: MAXIMUM.
G.L.A. INVE$T RESEARCH is a private investment research publication. This newsletter is for informational purposes only and does not constitute financial advice. All investments carry risk. Past performance of private companies is not indicative of future public market results. Always conduct your own due diligence.
© 2026 G.L.A. INVE$T RESEARCH — All Rights Reserved