AI Startups to Watch (Sep 16): Hang Ten, Arcee & TypeSafe
AI startups to watch on Sep 16, 2026: Vishal Sikka's Hang Ten adds $53M, Arcee hits a $1B valuation, and TypeSafe exits stealth with $40M.
AI startups kept the money moving on September 16, 2026: a four-month-old enterprise services firm founded by a former Infosys chief closed an eight-figure seed extension, an open-weight model lab crossed a billion-dollar valuation, and a security agent showed how fast a production environment can fall.
Hang Ten Systems: the IT services stack gets a new challenger
Hang Ten Systems is a Palo Alto enterprise AI services company founded in May 2026 by Dr. Vishal Sikka, the former Infosys CEO and ex-SAP product chief. It raised another $53 million five weeks after an initial $32 million seed round, taking total funding to $85 million.
Temasek's early-stage platform Xora led the extension, with Mayfield, Aramco Ventures, Intel CEO Lip-Bu Tan and Yahoo co-founder Jerry Yang. Sikka says Hang Ten signed a mission-critical software build 25 days after a first customer meeting, works across 21 large enterprises, and chases eight-figure deals. (Source: TechCrunch)
The pitch: AI code generation pushes the marginal cost of enterprise software toward zero, turning Accenture, TCS and Sikka's old employer Infosys from execution partners into the thing being replaced.
Arcee AI: open weights reach a billion-dollar valuation
Arcee AI is a US-based developer of open-weight frontier models. It closed a Series B led by Vista Equity Partners, valuing the company above $1 billion, with Cambium Capital, Emergence Capital, Hitachi, IAG and Microsoft's M12 fund participating.
Its flagship Trinity Large is a 400-billion-parameter sparse mixture-of-experts model that activates roughly 13 billion parameters per token, and Arcee says it trained its entire 2025 lineup for around $20 million. (Source: Arcee AI)
A billion-dollar valuation for open-weight models undercuts the claim that only closed labs can afford frontier capabilities, and it hands enterprises a version they can self-host.
TypeSafe AI: a model that does not chat
TypeSafe AI is a research lab that left stealth with $40 million in seed funding led by DCVC, reported by Forbes to carry a roughly $200 million valuation. Its team includes Diogo Almeida, an ex-OpenAI researcher who worked on RLHF, the technique that turned language models into usable assistants.
TypeSafe's System One family is built around Jev, a model that emits typed probabilistic decisions instead of text, trained with reinforcement learning from community disagreement. The company claims Jev is 193 times faster and 444 times cheaper per decision than a language model tool-call. (Source: The Register)
If decisions are that cheap, a large share of agent workloads stops paying per-token prices to frontier labs.
Noetive: sensors plus self-improving AI for physical work
Noetive is a San Francisco company that calls itself an AI intelligence of record platform for the physical economy. It raised $41 million in seed funding led by Eclipse, with Craft Ventures, the Westly Group, Factory and Gigascale Capital.
Chief executive and co-founder Amir Frenkel says the platform pairs a self-improving reasoning system with proprietary multi-modal sensing pods that perceive an environment, tie into enterprise tools, and coordinate people, machinery and inventory in real time. (Source: FinSMEs)
Physical operations software is dominated by systems of record that describe the past. Sensing pods plus self-improving models aim at the decision itself.
BackOps: an AI resolution layer for supply chains
BackOps is a San Francisco company selling what it calls an AI resolution layer for companies that make or move physical goods. It raised $42 million in Series B funding led by Insight Partners, with Theory Ventures and Construct Capital, and serves clients in grocery, retail, logistics and distribution.
Led by chief executive Sean McCarthy, BackOps captures internal operational processes, ships prebuilt connections to carrier systems, and executes approved work across enterprise systems, teams and carrier networks. (Source: FinSMEs)
It targets the exception handling that keeps freight moving, the manual work freight brokers and 3PLs bill for today.
Strix AI: an agent that found a production token in 25 minutes
Strix AI is a security startup selling autonomous offensive security agents. In a controlled test against Baseten, a model inference platform, a Strix agent located a hardcoded GitHub personal access token inside a container image within 25 minutes and reached private code repositories with administrative rights.
Baseten runs Harbor, its internal deployment platform for customer model containers, so the exposure reached the same supply chain Baseten sells to customers. It remediated the same day. (Source: Strix AI)
Penetration testing is sold as multi-week engagements. An agent that finds a repo-admin token in half an hour resets the price of that category.
Startup Snapshot
| Startup | What They Do | Latest Round | Valuation | Key Differentiator |
|---|---|---|---|---|
| Hang Ten Systems | Enterprise AI services and software builds | $53M seed extension | Not disclosed | Founder-led AI-native IT services model |
| Arcee AI | Open-weight frontier language models | Series B, Vista Equity | Above $1B | 400B-parameter sparse MoE, low training cost |
| TypeSafe AI | Decision models for machines | $40M seed, DCVC | About $200M reported | Jev emits typed decisions, not chat |
| Noetive | Sensing plus reasoning for physical operations | $41M seed, Eclipse | Not disclosed | Self-improving AI with sensing pods |
| BackOps | AI resolution layer for supply chains | $42M Series B, Insight | Not disclosed | Executes work across carrier networks |
| Strix AI | Autonomous offensive security agents | Undisclosed | Not disclosed | Agent-driven penetration testing |
Frequently Asked Questions
What are the top AI startups to watch today?
The top AI startups today are Hang Ten Systems, Arcee AI, TypeSafe AI, Noetive, BackOps and Strix AI. Hang Ten added $53 million for AI-native enterprise software delivery, Arcee crossed a billion-dollar valuation with open-weight models, and TypeSafe emerged with decision models that skip language model tool calls entirely.
Why did Hang Ten Systems raise again so quickly?
Hang Ten raised again because customer demand arrived faster than its original plan assumed. The $53 million extension landed five weeks after its $32 million seed round, led by Temasek's Xora, after the company signed multimillion-dollar enterprise contracts and built a pipeline across 21 large enterprises.
What makes Arcee AI's models different?
Arcee AI ships open-weight frontier models that enterprises can host themselves. Its Trinity family uses sparse mixture-of-experts designs, so only a fraction of parameters run per token, and the company says it trained its 2025 lineup for roughly $20 million, a fraction of what closed labs spend.
The Bottom Line
Three themes run through today's rounds. Enterprise services are being rebuilt around AI-native delivery, which is why a four-month-old firm can raise $85 million. Open-weight model builders now command billion-dollar valuations, which pressures per-token pricing. And autonomous security agents are becoming a category of their own, not a feature.
Yesterday's AI startups digest covered the physical AI wave that Noetive and BackOps are now chasing, and you can follow the series on the AI Startups tag page.
Related reading
Deeper context from the archive on the companies and technologies in this story:
- AI Startups to Watch (Sep 18): Crusoe, Emulate & Arcee
- Starcloud's .3B orbital data centers, a 1B inference chip bet, and robot models that learn from a single demo: AI Startups to Watch for August 23, 2026
- AI Startups to Watch (Sep 12): Cohere, XDOF & Kinetix
- AI Startups to Watch (Sep 11): Positron, Cognition & Fluidstack
Building an AI agent, automation, or backend workflow? AsaanKaroobar builds custom AI solutions, apps, and digital experiences. Get in touch →
