The first week of March 2026 was not supposed to happen. Venture capital has spent two years preaching discipline: fewer bets, higher bars, slower checkbooks. Then, in roughly 72 hours, investors wrote more than $2.6 billion in checks to a handful of US startups. At least three of those rounds cleared $500 million. The market did not get the memo.
Two themes are pulling in capital at a scale that looks less like a funding round and more like a capital project: space infrastructure and AI hardware. Four companies defined the week.
Stop 1: Ayar Labs, the Plumbing AI Actually Needs
Most people have never heard of co-packaged optics. That is about to change.
Ayar Labs, based in Santa Clara, closed a $500 million Series E on March 3, backed by Nvidia, AMD, MediaTek, Neuberger Berman, ARK Invest, and the Qatar Investment Authority. The round values the company at $3.75 billion and brings total funding to $870 million. CEO Mark Wade called it the company’s last private round before an IPO.
Ayar makes optical interconnects, essentially chips that move data using light instead of copper wire. In an AI data center, the bottleneck is not the GPU. It is moving data between GPUs fast enough to keep them busy. Copper cables cannot keep up at the bandwidth levels that next-generation AI models demand, and Ayar’s co-packaged optics replace those copper links with something faster and substantially more power-efficient. The IEA’s 2025 Energy and AI report puts the scale of the problem in numbers: electricity consumed by AI-accelerated servers is growing at 30% per year, and global data center power demand is on track to nearly double by 2030, reaching 945 TWh annually. Efficiency at the interconnect layer is no longer an engineering preference. It is a cost question.
The investment thesis is simple. Nvidia’s customers are committing to volume ramps in 2028, and they need the interconnect stack validated two years before that. Ayar has a short window to become that standard. The $500 million funds the production scale and ecosystem lock-in to make it happen. Investors are not betting on a science project. They are buying a supply-chain position inside the next wave of AI infrastructure.
Stop 2: Sierra Space, Where Defense Contracts do the Convincing
Sierra Space closed a $550 million Series C on March 5, led by LuminArx Capital Management, with General Atlantic and Coatue among returning investors. The round values Sierra at $8 billion, up from $5.3 billion in 2023. Total capital raised since 2021 now exceeds $2 billion.
Sierra builds satellites and spacecraft for national security customers. Its Dream Chaser spaceplane (a reusable winged vehicle that launches on a rocket and lands on a runway) is scheduled for its demonstration flight in late 2026. The company also holds a $450 million contract to build satellites for an unnamed national security customer and a Space Development Agency contract worth up to $740 million for missile tracking satellites.
Those government contracts are the real story. They give investors long-term revenue visibility that an equity narrative alone cannot provide. A VC writing a $100 million check to a pure commercial-space startup is making a market bet. A VC writing into Sierra Space is co-investing alongside the US Department of Defense. The government already did the diligence. Private capital is riding that validation.
For investors who have been burned by commercial space companies that ran out of money before reaching orbit, a funded program of record genuinely changes the risk calculation. BryceTech’s Start-Up Space 2025 report, the most comprehensive annual tracker of private space investment, put total 2024 sector funding at $7.8 billion globally, with $4 billion going to US companies. Sierra and Vast together claimed more than a sixth of that annual US total in a single week.
Stop 3: Vast, Betting on What Comes After the ISS
On the same day Sierra announced its round, Vast closed $500 million ($300 million in equity, $200 million in debt), led by Balerion Space Ventures, with the Qatar Investment Authority and Mitsui among the participants. The company now has more than $1 billion in total funding and over 1,000 employees.
Vast is building private space stations. Haven-1, its first commercial station, is set to launch in 2027. A larger successor called Haven-2 is designed to keep humans continuously in low Earth orbit after the ISS retires later this decade. In February 2026, NASA selected Vast for its sixth private astronaut mission to the ISS, a meaningful proof point ahead of Phase 2 competition for the next US national space station contract. Worth being direct about one thing: Vast did not win the Phase 1 contract. This $500 million is partly a comeback bet.
The investment case is about timing more than technology. The ISS is approaching end-of-life. NASA needs a commercial successor, and Congress is not going to fund a government-owned replacement at the same scale. That creates an opening where private capital builds infrastructure the federal government will essentially lease, a model that looks more like airports than satellites. Whether investors are right depends almost entirely on whether Vast wins Phase 2.
Stop 4: Science Corp, the Outlier Worth Studying
Science Corp raised $230 million in a Series C on March 5, smaller than the others but worth understanding. The company, founded by Neuralink co-founder Max Hodak, is commercializing PRIMA: a retinal implant smaller than a grain of rice that restores functional vision in patients with advanced macular degeneration. Lightspeed, Khosla Ventures, Y Combinator, and IQT backed the round. Post-money valuation: $1.5 billion.
Science Corp has nothing to do with space or AI infrastructure. What it shares with the other three companies is a hardware bet on a specific physical bottleneck, in this case the gap between what existing neurotechnology can do and what the nervous system actually requires. The company has submitted its CE mark application in Europe and expects approval by mid-2026, which would make it the first BCI company with a product in market.
PRIMA is a regulated medical device with a defined patient population, a real reimbursement pathway, and a near-term revenue date. Hard tech plus near-term commercialization: that is the combination serious investors are paying up for in 2026. Not hype. A timeline.
What this Week Tells Founders
Capital is concentrating, not growing
This $2.6 billion did not come from a bigger pool of VC money. It came from a market that increasingly prefers writing one large check to a single winner over seeding ten companies and watching what survives. Crunchbase’s full-year 2025 analysis found that more than a third of all global VC went to just 68 companies that raised $500 million or more, up from 24% in 2024. General Catalyst is reportedly raising $10 billion. Founders Fund is closing a $6 billion fund. Andreessen Horowitz raised $15 billion in January. The money is piling up at the top of the funnel and moving in concentrated bursts to companies that fit specific narratives.
For founders raising a seed or Series A in 2026, this creates two practical problems. The bar for narrative fit is higher than it was. Being in AI is not enough; you need to be solving a specific hardware, infrastructure, or platform constraint that the biggest rounds are not already funding. And top-tier investor attention is increasingly consumed by their largest portfolio companies. Getting into a meeting is harder when the partner just deployed $500 million.
Which narratives are overfunded
AI application software is crowded. Too many companies are building on top of foundation models with similar go-to-market motions and thin differentiation. Investors still look at these deals, but valuations have compressed and the traction bar is considerably higher than it was 18 months ago.
Consumer hardware is largely unfunded after a string of high-profile failures. Late-stage climate tech without a contracted government or utility offtake is a similarly hard sell right now.
Which narratives are underfunded
Physical infrastructure with long economic lives and a government or enterprise anchor is where serious capital is moving: AI compute infrastructure, defense-adjacent space, next-generation semiconductor packaging, regulated healthcare hardware like PRIMA. These are not fast flips. They are bets on 7- to 12-year capital cycles.
There is also an underappreciated market in unglamorous B2B software that sits inside these infrastructure stacks: logistics tooling, compliance systems, data management for defense contractors. These companies rarely make the headline rounds, but they get acquired. The companies writing acquisition checks in 2028 will be companies like Sierra Space and Vast, looking to own their supply chains.
The Bottom Line
This week was not an accident. The companies that raised $500 million or more in early March 2026 all have hardware at the center of their product, a government or major enterprise anchor in their revenue plan, and a specific physical problem that software alone cannot solve.
The VC market is selective, not cautious. The money is there. The question is whether your company fits the narrative pulling capital right now, or whether you are still trying to raise in a category that already had its moment.
