Updated May 2026
TL;DR
The biggest AI funding rounds this week (week ending 22 May 2026) pulled in $4.15bn across 10 US deals, with frontier AI, AI infrastructure and hardware taking the bulk.
Hark's $700m Series A at a $6bn valuation, backed by Nvidia, AMD, Qualcomm and Intel, reopens the personal AI device narrative just as Humane and Rabbit's corpses are still warm.
Modal Labs went from $60m ARR to $300m in 8 months. Enterprise AI coding is now a real budget line, not a pilot.
Mercury's $200m Series D and recent OCC bank charter approval confirm the default banking stack for AI-native founders: 1 in 3 US startups now banks with them.
If you are launching this summer, your story has to sit inside one of three live investor narratives: personal AI hardware, AI-native infrastructure, or applied vertical AI in defence, healthcare or retail. Anything else fights the news cycle.
Why this week's funding mattered
Eight of the top 10 US rounds were AI-adjacent. The other two were a $1.5bn medical device round and a fintech that banks AI startups. Even the deals without "AI" in the headline were downstream of the AI boom.
This is the cleanest week in months for reading where capital is actually moving, versus where the headlines say it is. The signal: investors are putting real cheques behind physical AI again, frontier labs are still being treated as a land grab, and the picks-and-shovels layer (compute, edge infra, banking) is being aggressively re-priced.
For founders preparing a launch, this is not abstract. Each of these deals opens a narrative window. Land your launch inside one and you get press coverage, investor inbound, and the right kind of competitive comparison. Miss the window and you are explaining your category from scratch to a room that has already moved on.
The four rounds that reset the narrative this week
Hark: $700m Series A and the second wave of AI hardware
Hark raised $700m at a $6bn valuation, led by Parkway Venture Capital with Nvidia, AMD Ventures, Intel Capital, Qualcomm Ventures, ARK Invest, Salesforce Ventures, Greycroft and Brookfield piling in (TechCrunch coverage here).
The detail most coverage skipped: Hark is Brett Adcock's third startup after Figure AI and Archer Aviation. That is the entire pitch. Adcock has shipped hardware once with Archer and is mid-flight with humanoid robots at Figure. Investors are not really buying Hark's roadmap. They are buying the founder's track record of getting physical product out the door at scale.
Hark currently runs 70 staff and a data centre stacked with Nvidia B200 GPUs. First product ships this summer.
Founder takeaway: the "personal AI device" category just got reopened a year after Humane sold to HP for $116m and Rabbit's R1 stalled out. The narrative window is open for roughly 3 to 6 months. If you are building anything adjacent (companion AI, on-device agents, AI peripherals) you need to be visible inside that window or you will look late by autumn.
Modal Labs: $355m Series C and proof enterprise AI coding is real revenue
Modal raised $355m at a $4.65bn valuation, led by General Catalyst and Redpoint with Bain Capital Ventures, Menlo and Accel joining. The standout sits inside the announcement: ARR went from $60m in September to $300m at close. That is a 5x in 8 months (Modal's own write-up).
CEO Erik Bernhardsson told Reuters the initial tranche closed at a $2.5bn valuation and the round was reopened to absorb inbound demand at $4.65bn. That is a tell about both the market and Modal's specific position. Enterprise customers running AI-generated code at production scale need sandboxes, GPU access and serverless primitives, and there is no clear category leader yet.
Founder takeaway: if your AI product touches code generation, agent execution, or any "AI does the work" workflow, you now have a buyer-side comparison. Modal is the reference customer your enterprise buyer already knows. Lean on it in the pitch.
Decart: $300m and the world-model land grab
Decart raised $300m at a $4bn valuation, led by Radical Ventures with Nvidia, Sequoia, Benchmark, Adobe Ventures and Toyota Ventures alongside. Andrej Karpathy, Michael Eisner and the Nintendo family came in as angels. Amazon joined as a strategic customer (Decart announcement here).
Karpathy on the cap table is the signal worth reading. He is selective. His name on a foundation-model bet outside the big three (OpenAI, Anthropic, xAI) tells you the frontier lab category is widening, not consolidating. Foundation model funding doubled YoY in Q1 2026, and the door is not closing.
Founder takeaway: there is still room for vertical world models, real-time generative systems and inference-optimised models. The pitch that wins right now is "we are not competing with OpenAI, we are doing the thing OpenAI cannot ship."
Mercury: $200m, $5.2bn valuation, and a bank charter
Mercury raised $200m at $5.2bn from TCV, with a16z, Coatue, CRV, Sequoia, Sapphire and Spark returning. The numbers behind the round: $650m ARR, profitable on GAAP for 4 consecutive years, 300k customers, 1 in 3 US startups banking with them, and applications up 2.5x year-on-year in Q1 (CNBC coverage here).
The actual headline is the OCC bank charter approval in April. Mercury went from "neobank with a BaaS partner" to "actual bank" inside 14 months.
Founder takeaway: if you have not switched to Mercury yet, you are increasingly the odd one out in fundraising conversations. Investors check what you bank with. It is a small signal that telegraphs whether you are inside the modern AI startup playbook or running an old one.
Where is AI venture capital flowing in 2026
Three places, in order of capital concentration:
Frontier labs. Q1 2026 alone saw OpenAI, Anthropic, xAI and Waymo close $188bn between them, 65% of global venture investment for the quarter (Crunchbase Q1 data here). Decart's $300m extends the pattern to the second-tier frontier labs.
AI infrastructure and developer tools. Modal, Armada ($230m for edge AI), Exa ($250m for AI-native search). Buyer demand is real, ARR multiples are holding, and the category is not yet consolidated.
Physical and applied AI. Hark for consumer hardware, Amca for aerospace, Radar for retail RFID plus AI. Anything that puts AI into a physical surface or workflow is being bid up because pure software AI is now considered crowded.
If your launch narrative does not sit inside one of these three buckets, you have a positioning problem before you have a PR problem.
What AI founders should do this week
Audit your launch narrative against the three live stories. If you cannot frame your product in 30 seconds as either personal AI hardware, AI infrastructure, or applied vertical AI, you are pitching into the wrong news cycle. Rewrite the deck this week.
Build a Hark comparison line. Reporters covering personal AI for the rest of 2026 will reference Hark the way they referenced Humane in 2024. Decide now whether you are "the Hark for X" or "what Hark gets wrong" and own the comparison before someone else does.
Use Mercury's customer count as social proof. If you bank with Mercury, say so on your About page. It is a 2-second signal to investors and operators that you are in the modern startup stack.
Pitch the secondary publications now. The Wall Street Journal will not cover your $4m seed. Sifted, The Information, Tech Funding News and Newcomer will. Their inboxes are emptier than TechCrunch's and their readers are the ones who write the cheques.
Plan around the narrative half-life. The Hark personal AI hardware window closes when the device ships and reviews land. That gives you roughly until early September. After that, "personal AI device" becomes a saturated comparison and you need a new wedge.
The rest of this week's top 10
MiRus, $1.5bn corporate round (healthcare). Boston Scientific took a 34% stake. Strategic capital is back in medtech, and the deal pulled the headline slot away from AI for the week.
Amca, $300m Series B (aerospace and defence). Caffeinated Capital led, a16z and Construct Capital joined at a reported $1bn+ valuation. Defence tech is still the second-hottest non-AI category.
Exa, $250m Series C (AI search). a16z led at a $2.2bn valuation. AI-native search infrastructure for agents and LLM applications.
Armada, $230m Series B (edge AI). 8090 Industries, BlackRock and Overmatch Ventures led at $2.2bn. Edge compute for remote and industrial environments.
Radar, $170m Series B (retail tech). Gideon Strategic Partners and Nimble Partners led at a $1bn unicorn valuation. Overhead RFID plus AI for brick-and-mortar inventory. Already deployed in American Eagle and Old Navy across 1,400+ stores.
Farther, $150m Series D (wealth management). General Atlantic led. Tech-enabled wealth advisory. New unicorn, valuation undisclosed.
FAQ
What were the biggest AI funding rounds this week (ending 22 May 2026)?
Hark led the AI deals at $700m Series A and a $6bn valuation. Modal Labs raised $355m Series C at $4.65bn. Decart raised $300m at $4bn. Exa raised $250m Series C at $2.2bn. Armada raised $230m Series B at $2.2bn. Across the top 10 US rounds, roughly $4.15bn was deployed in a single week.
Who led Hark's $700m Series A?
Parkway Venture Capital led, with Nvidia, AMD Ventures, Intel Capital, Qualcomm Ventures, ARK Invest, Salesforce Ventures, Prime Movers Lab, Greycroft, Brookfield and Tamarack Global participating. Hark was founded by Brett Adcock, also the founder of Figure AI and Archer Aviation.
Why is "physical AI" funding accelerating again in 2026?
Investors are betting the next wave of AI value capture sits at the hardware and physical-system layer, not pure software. Frontier labs absorbed 80% of Q1 2026 VC dollars in AI, which has crowded out the software-only category. Capital is now rotating into AI hardware, robotics, edge compute and applied AI in defence, healthcare and retail.
What does Mercury's $200m Series D mean for AI startups?
Mercury now banks roughly 1 in 3 US startups and just secured OCC bank charter approval. For AI founders, it confirms Mercury as the default banking choice for the modern startup stack, and signals to investors that you are inside the current playbook rather than running a legacy one.
Which AI funding categories should founders position around for the rest of 2026?
Three live narratives are pulling press and capital: personal AI hardware (Hark, Decart's world models), AI-native infrastructure (Modal, Armada, Exa), and applied vertical AI in defence, healthcare and retail (Amca, MiRus, Radar). Launches mapped onto one of these get coverage and inbound. Launches outside them are explaining the category from scratch.
Where is AI venture capital flowing in 2026?
Capital is concentrating in three layers. Frontier labs absorbed roughly $188bn in Q1 2026 across OpenAI, Anthropic, xAI and Waymo. AI infrastructure and developer tools are the second concentration, with Modal Labs hitting $300m ARR as proof of enterprise demand. Physical and applied AI is the third, with Hark, Amca and Radar pulling capital into hardware-led categories.
Get this every Monday
We track the funding rounds that move the AI narrative each week and turn them into launch positioning for the founders we work with. Talk to Ignita before your next launch.
