Giga Omni Ventures

Giga Omni Ventures was founded in 2018 by former Google DeepMind researcher Dr. Priya Sundaram and ex- Sequoia partner Marcus Chen after they spent three years watching AI transition from academic curiosity to infrastructure layer. Their thesis was simple: the next trillion-dollar companies would be built on AI-native infrastructure, not companies with AI bolted on top. With a first fund of $280M anchored by sovereign wealth capital and a handful of iconic angel investors, GOV has backed 34 companies across three funds.

The firm's name reflects its conviction: 'Giga' for the scale of compute and data driving modern AI, 'Omni' for the horizontal applicability of foundation models across industries. Sundaram and Chen built GOV specifically to write early checks ($1M-$5M pre-seed and seed) into companies building the enabling layer, not the application layer.

Today GOV manages $1.1B across three funds and has produced four unicorn exits, including neural memory startup Cortexica (acquired by Snowflake for $840M) and distributed training infrastructure firm LatticeDB (IPO'd at $4.2B market cap in 2025). The firm operates from offices in San Francisco, London, and Singapore.

Understanding Giga Omni Ventures's approach is essential for founders building in deep tech. Whether you are at the pre-seed stage with a research-heavy team or raising a Series A for your AI infrastructure startup, this guide will help you understand how to position your company for GOV's investment committee.

In addition to capital, Giga Omni Ventures provides portfolio companies with access to compute partnerships with major cloud providers, a network of academic advisors from top AI labs, and operational support for scaling engineering teams. The firm is known for being unusually helpful with early hiring, often making introductions to senior engineers and product leaders who have passed on joining larger AI labs.

GOV's investment thesis rests on a structural observation: every platform shift in computing has produced a new 'picks and shovels' layer. The gold rush around LLMs has already created enormous demand for infrastructure that makes training, serving, and fine-tuning models cheaper and more reliable. GOV specifically avoids pure application-layer plays, preferring companies whose revenue comes from enabling AI rather than competing with the models themselves.

The firm groups its interest into four concentric rings. The innermost ring is compute orchestration and distributed training—companies building the plumbing for multi-GPU and multi-node training at scale. The second ring is data infrastructure for AI: vector databases, training data pipelines, synthetic data generation, and data quality tooling. The third ring is AI safety and alignment tooling, an area Sundaram personally champions. The outermost ring is what Chen calls 'AI-first verticals'—industries where the core business logic depends on AI in a way that cannot easily be replicated by incumbents.

GOV prefers to invest at the infrastructure layer but has shown flexibility when the application is deeply integrated with proprietary data moats or requires specialized hardware. Companies like Axiom Photonics (photonic compute for AI inference) and Meridian Privacy (federated learning infrastructure for regulated industries) demonstrate this nuance.

The firm typically leads or co-leads rounds and takes a board seat in most investments. GOV's model is deeply collaborative: the firm encourages portfolio companies to share learnings and, in several cases, has facilitated partnerships between portfolio companies who were natural上下游 complements.

GOV has a stated preference for 'founder-market fit above all else.' In practice this means they have backed PhDs with no business experience and seasoned operators with no ML background, as long as the combination of person and problem is exceptional. Their deal memos frequently reference the contrarian signal: if the obvious incumbents aren't already building this, why not?

Dr. Priya Sundaram serves as GOV's Managing General Partner and leads the firm's AI safety and alignment portfolio. Before co-founding GOV, Sundaram spent seven years at Google DeepMind working on alignment research and large-scale distributed training. She holds a PhD in computational neuroscience from MIT and was a Rhodes Scholar. Sundaram has published 23 peer-reviewed papers, with 14 focused on RLHF and reward modeling. At GOV she maintains an active research collaboration with the Stanford HAI lab and serves on the technical advisory board of the Partnership on AI.

Marcus Chen is GOV's General Partner focused on infrastructure and compute. Chen spent nine years at Sequoia Capital where he led investments in infrastructure companies totaling over $2B in aggregate value. Prior to Sequoia he was a product manager at Cisco's data center group and an early engineer at Mesosphere (acquired by D2iQ). Chen graduated from Carnegie Mellon with a BS in computer science and an MBA from Wharton. He is known in the Bay Area for writing detailed technical diligence notes and for being unusually accessible to seed-stage founders who need feedback.

The two met during Chen's investment in a Sundaram-affiliated spin-out from DeepMind in 2017. Chen led Sequoia's seed check; Sundaram was the technical founder. After the acquisition, they stayed in touch and spent six months conducting 80 founder interviews before deciding to start GOV together.

GOV also has two operating partners: Yuki Tanaka (formerly VP Engineering at Cerebras, focused on hardware diligence) and Dr. Amara Osei (formerly at DARPA's I2O, focused on government and defense AI contracts). Both serve as resource anchors for portfolio companies rather than deal leads.

In 2025 GOV deployed $180M across 14 new investments and 6 follow-on rounds. The firm has remained active in the current market environment, which Chen describes as 'selective but not conservative—we still write the same number of checks, we just see more quality deal flow.' Sundaram notes that the AI safety category has seen a marked increase in deal quality as enterprise buyers have started treating AI governance tooling as a compliance requirement rather than a nice-to-have.

The firm's 2025 investments included a $4M pre-seed in ZeroRupture (automated red-teaming for LLM applications), a $2.5M seed in Cascade Protocol (distributed inference optimization for edge deployment), and a $12M Series A in Ironclad AI (a developer-first MLOps platform targeting regulated financial services). All three deals were done without co-investors at the lead, a pattern Chen says the firm is deliberately maintaining to preserve ownership.

GOV's geographic expansion to Singapore in 2024 has opened a new deal flow channel in Southeast Asian AI infrastructure, where the firm has made three investments including a $3M seed in Nusantara Compute (cloud GPU exchange for the archipelago market) and a Series A follow-on in Meridian Privacy.

The firm closed its third fund, a $620M vehicle, in Q3 2025, anchored by a $150M commitment from a North American public pension fund and participation from three major university endowments. The fund will maintain the same check size ranges as Fund II.

Cortexica built a neural memory layer that allows foundation models to retain and recall long-horizon context without catastrophic forgetting. GOV led their $4M seed in 2020 and co-led their $22M Series A in 2021. The company was acquired by Snowflake in 2023 for $840M, representing GOV's largest exit to date.

LatticeDB developed a distributed vector indexing system optimized for billion-scale embedding databases with sub-10ms query latency. GOV led a $3M seed in 2021. The company raised a $45M Series B in 2024 and IPO'd in early 2025 at a $4.2B market cap, giving GOV a 90x return on its seed position.

Meridian Privacy provides federated learning infrastructure for banks and insurers that need to train on sensitive customer data without crossing compliance boundaries. GOV led their $6M Series A in 2022 and participated in the $35M Series B in 2024. Annual Recurring Revenue crossed $40M in Q4 2025.

Axiom Photonics is developing photonic tensor processors for AI inference workloads, targeting the datacenter market with 40x energy efficiency gains over NVIDIA's H100. GOV led a $9M Series A in 2023. The company recently emerged from stealth with a partnerships with two Tier 1 cloud providers.

Cascade Protocol builds compiler-level tooling that optimizes distributed inference across heterogeneous edge devices, dramatically reducing the memory and compute footprint of running frontier models on-device. GOV led a $2.5M seed in 2025.

Ironclad AI is a developer-first MLOps platform purpose-built for regulated environments, with deep integrations into financial services workflow tooling. GOV led a $12M Series A in 2025. The company is currently at $18M ARR benchmarks with 3x year-over-year growth.

GOV evaluates companies on four dimensions: technical differentiation, market timing, team composition, and defensibility. The firm looks for companies with genuine IP moats—patents, proprietary datasets, or architectural innovations that cannot be replicated in under 18 months by a well-funded team.

Market timing is critical. Sundaram describes her framework as 'three years too early is indistinguishable from wrong.' GOV has passed on several technically excellent companies that were addressing markets that, in Chen's words, 'needed another model generation to become real.' The firm tracks cloud GPU pricing trends, fine-tuning cost curves, and enterprise procurement cycles as timing signals.

Team composition at GOV is evaluated with less rigidity than at most deep tech funds. The ideal founding team has at least one person who has personally built and shipped the thing the company is selling, but GOV does not require a PhD or prior founder experience. What the firm does require is intellectual honesty—a track record of updating beliefs based on evidence.

Defensibility is assessed on a case-by-case basis. GOV generally prefers businesses with natural network effects, proprietary data loops, or switching costs over pure technical moats, because the latter erode faster in a world where foundational model capabilities improve rapidly.

The firm explicitly avoids pure SaaS plays (software-only infrastructure without an AI component), consumer applications at seed stage, and companies building on top of a single cloud provider in a way that creates platform dependency.

GOV has a published 'Anti-Portfolio' on its website—a list of companies they passed on that went on to succeed. Chen says this is meant to signal intellectual humility and to remind the team that the best investments often look wrong at first pass.

The highest-conversion path into GOV is a warm introduction from one of their portfolio founders, an academic advisor on GOV's technical advisory board, or a trusted investor who has worked with Chen or Sundaram previously. The firm tracks source of inbound and measures conversion rates by channel: introductions from portfolio founders convert to term sheets at roughly 4x the rate of cold inbound.

The second-best channel is through GOV's academic network. Sundaram maintains close ties to Stanford HAI, MIT CSAIL, and Berkeley's BAIR lab. PhD researchers at these institutions who are considering spinning out their work frequently get introductions through these channels, and several GOV investments have originated from Sundaram recognizing a paper and reaching out directly to its authors.

Cold email to the firm goes to a dedicated deal sourcing team. Chen has said publicly that his personal inbox sees about 800 emails per week and that cold inbound founders have about a 2% response rate. The firm's website has a pitch submission form that routes to the same team. Founders who submit through the form with a clear articulation of their technical differentiation and a demonstration of relevant domain expertise tend to get responses.

GOV attends NeurIPS, ICML, and MLSys annually and uses conference attendance as a sourcing channel. The firm also hosts an annual 'Deep Tech Dinner' in San Francisco each October that brings together 80 portfolio founders, potential investees, and advisors for a single evening of curated conversation.

For founders who have been referred by a portfolio CEO, GOV's typical process is: introductory call (30 minutes), technical deep-dive (60-90 minutes with an operating partner), partnership review (60 minutes with both GPs), and final partnership committee (30 minutes). The full process typically runs 6–8 weeks from first contact to term sheet.

While GOV invests in early-stage deep tech companies, they expect founders to have a solid handle on their financials. For infrastructure companies this includes understanding your gross margin trajectory as you scale, your cloud partner dependencies, and your customer concentration risk.

Many technical founders underestimate the importance of financial clarity during due diligence. GOV's investment committee will scrutinize your SaaS unit economics even at the seed stage, particularly for companies with hardware or cloud infrastructure components where COGS are non-trivial.

Working with a fractional CFO can significantly improve your chances of securing funding. Professional financial guidance helps you build accurate projections, prepare investor-ready financials, and confidently answer the operational questions that GOV's partnership committee asks in the final stages of diligence.

Our team has helped numerous deep tech and AI infrastructure companies raise venture capital and would be happy to discuss how we can support your fundraising efforts. From pitch deck financials to comprehensive financial models with cloud cost assumptions built in, we ensure you are prepared for the investment process.

Financial projections should be grounded in evidence and realistic about cloud cost curves. GOV will challenge your assumptions on gross margin—especially for companies that rely on a single cloud provider for compute. Be prepared to discuss your pricing power and how you plan to reduce COGS as you scale.

Understanding your key performance indicators is essential when pitching to GOV. The firm tracks different metrics than typical SaaS investors: training efficiency (TFLOPS per dollar), inference cost per query, and data pipeline reliability are often more relevant than traditional ARR benchmarks growth rates.

Whether you are preparing to pitch Giga Omni Ventures or other deep tech VCs, having professional financials can set you apart from the competition. Our team has helped companies build the financial models that investors like GOV expect to see in due diligence.

Exploring other venture capital firms? Our comprehensive collection of VC firm reviews covers hundreds of investors across all stages and sectors.

Each review provides detailed information about investment criteria, portfolio companies, and tips for securing funding. Whether you are looking for seed-stage investors or growth equity firms, you will find valuable insights in our VC firm guides.

Finding the right investor for your startup is crucial to your success. Take the time to research potential investors and understand their investment thesis before reaching out.

Our guides cover major venture capital firms as well as emerging managers that may be a better fit for your company's specific needs and stage.