Elad Raz: NextSilicon's Challenge to Nvidia

Founder and CEO, NextSilicon · The Israeli reconfigurable-compute chip company now deployed inside Sandia National Laboratories · $1.6B reported valuation, 2024 Series E.

Elad Raz founded NextSilicon in 2017 to build a reconfigurable-compute architecture that adapts its silicon to the application at runtime — a direct architectural challenge to Nvidia's fixed-GPU dominance of the high-performance computing and AI-training markets. NextSilicon's Maverick processor is now deployed inside Sandia National Laboratories under a US Department of Energy contract, a rare direct US-federal-HPC placement for an Israeli chip startup. The company's 2024 Series E, reported at $1.6 billion valuation, positions Raz as one of the very small number of Israeli founders operating at the frontier of the semiconductor architecture layer that AI compute now depends on.

Background

Elad Raz's pre-NextSilicon career sits inside the Israeli signals-intelligence-to-industry pipeline that Unit 8200 and the broader IDF technical corps have supplied to Israeli industry for four decades. His technical background is in silicon architecture and high-performance-computing systems, developed through Israeli academic and industrial channels before he founded NextSilicon at 32.

The Maverick architecture

The Maverick processor is NextSilicon's flagship product: a data-flow reconfigurable-compute accelerator that observes the target application at runtime, generates an optimized data-flow graph for that application, and reconfigures its own silicon to execute that graph natively. The architectural claim is that this closes the abstraction gap that fixed-instruction-set architectures (CPUs, GPUs) leave open — the gap between the general-purpose instruction set and the specific application's compute pattern.

The Maverick's target workloads are scientific simulation (climate, molecular dynamics, fluid dynamics), graph analytics (network analysis, financial risk, national-security applications), and machine-learning inference at HPC scale. On reported benchmarks in the HPC community, Maverick has demonstrated substantial performance advantages over both CPU and GPU baselines on the specific workloads it is designed for.

The Sandia National Laboratories deployment

NextSilicon's placement inside Sandia National Laboratories — the US Department of Energy's flagship HPC and national-security research facility — is the single most consequential validation event in the company's history. Sandia is one of the world's most demanding HPC customers: its workloads span nuclear-weapons stockpile stewardship simulation, energy-grid modeling, and national-security AI. The DOE's national-lab procurement cycle is famously conservative and famously long — a placement at Sandia signals both technical validation and successful navigation of the US federal procurement process at a level of scrutiny that very few Israeli chip startups have cleared.

The 2024 Series E and the capital-formation profile

NextSilicon's 2024 Series E was reported at approximately $1.6 billion post-money valuation. The company's cumulative funding across prior rounds is reported in the range of $300 million-plus. Nvidia's total-cost-of-ownership advantage — CUDA ecosystem, developer tooling, installed base — represents an operational moat that architectural challengers must fund past, not merely fund up to.

Where NextSilicon sits in the Israeli semiconductor economy

Israel's semiconductor position is anchored by four large-scale operating names: Intel Israel; Tower Semiconductor; the NVIDIA Israeli engineering center (post-Mellanox acquisition, 3,000+ employees, the AI-compute networking-layer anchor); and a broader wave of Israeli-founded chip startups. NextSilicon sits inside that broader wave at the highest architectural risk-and-reward level.

Watch points

• Sandia deployment performance data — the first substantive independent-lab validation results as they become public.
• Additional US federal HPC placements — Oak Ridge, Argonne, Lawrence Livermore, Los Alamos as follow-on procurement targets.
• The commercial hyperscaler question — whether NextSilicon can convert the federal-lab validation into cloud-provider deployment.
• The Nvidia response — whether Nvidia's incremental CUDA and Grace-Hopper roadmap addresses the reconfigurable-compute market before NextSilicon reaches revenue scale.
• Israel's national semiconductor policy — whether NextSilicon receives strategic industrial support from Israeli government mechanisms.

Why Elad matters

The Israeli AI economy is now materially dependent on Nvidia. Every Israeli AI application company, every Israeli AI-security startup, every Israeli AI-defense contractor buys compute upstream from a single US corporate supplier. NextSilicon is the single Israeli-founded operating company positioned at the layer where that dependency could, in principle, be reshaped. The Sandia validation is not proof that the reshaping will happen. It is proof that the technical thesis has cleared federal-lab scrutiny.

Sources

NextSilicon company disclosures on funding rounds, product releases, and customer deployments. Sandia National Laboratories procurement announcements. Reporting in Calcalist, Globes, Haaretz, and The Marker on NextSilicon's commercial trajectory. Reuters, Bloomberg, and industry press coverage of the 2024 Series E.

Related Olam Coverage

See also: The Builders · Eyal Waldman (Mellanox / NVIDIA) · Olam AI Citation Share Index — 2026 Series · Israeli Semiconductor Citation Share Index 2026