Two recent funding rounds tell a single story founders pitching consumer AI wrappers should read carefully. Salience Labs raised a $125 million Series B for silicon photonics and optical circuit switching for AI data centers; Lumaril raised a $66 million Series A at a $400 million valuation for industrial AI digitizing operator knowledge. Different layers, same direction: capital is flowing to factories, machines, and physical AI infrastructure, not another chatbot skin.
Salience Labs: the physics of the data center
Salience Labs builds silicon photonics, including optical circuit switching (OCS), for AI data centers. Its $125 million Series B was led by Bullhound, with AMD Ventures and Intel Capital participating — a signal that chip giants see this technology as strategically necessary.
The logic is straightforward: AI training and inference move enormous data volumes between chips and racks, and copper interconnects hit limits on speed, power, and heat. Photonics — moving data as light — promises higher bandwidth at lower power. As data centers scale, the interconnect becomes a bottleneck, and whoever owns it owns a valuable piece of the stack.
This is deep-tech investing in its classic form: hard physics, long development cycles, and a payoff that compounds if the bet is right. A $125 million Series B at this stage suggests investors believe the technology is past the science experiment and into the commercialization window, with data center operators as the natural customers.
Lumaril: the knowledge walking out the factory door
Lumaril attacks a different, equally physical problem. Its industrial AI digitizes operator knowledge — the tacit expertise of experienced plant workers, the kind of know-how that lives in someone's head after thirty years on a line and vanishes when they retire. Its $66 million Series A, at a $400 million valuation, was led by Elad Gil, with Felicis, Koch Disruptive, Lux, and General Catalyst participating.
That investor list is a statement: Elad Gil rarely leads rounds that are not category-defining bets, Koch Disruptive brings the industrial operator's perspective, and Lux and General Catalyst bring deep-tech and enterprise scaling experience. Together they bet the retirement wave in skilled manufacturing is structural — every factory losing veteran operators is a customer for software capturing what they knew.
The pattern: physical AI, not wrapper AI
These rounds do not stand alone. Recent coverage has also included Foundry Robotics, which raised a $100 million Series A for autonomous factories, and Bedrock Robotics, which raised a $600 million Series C. Taken together, the pattern is unmistakable: the biggest checks in AI right now are going to companies that touch the physical world — factories, robots, chips, and the infrastructure that moves data between them.
There are good reasons for this rotation. First, consumer AI wrappers face brutal competition: a product built on someone else's model, with no proprietary data or distribution moat, can be cloned in weeks. Investors learned that lesson quickly.
Second, physical-world AI has natural moats. Access to factories, proprietary operational data, hardware expertise, and regulatory certification are hard to replicate; a startup deployed in fifty plants has something a new entrant cannot copy from a laptop.
Third, the customers can pay. Manufacturers, data center operators, and industrial conglomerates have real budgets and urgent problems — labor shortages, energy costs, throughput targets — and they renew software that demonstrably saves money or prevents downtime.
What this means for founders
If you are building in AI, the funding lane has moved. Three practical lessons follow from the Salience and Lumaril rounds.
Sell to a bottleneck, not a trend. Salience sells into a physical bottleneck — data movement in data centers. Lumaril sells into a demographic one — retiring expertise. Both start from a constraint the customer already feels. Founders should be able to name their customer's bottleneck in one sentence.
Proprietary access beats proprietary models. Neither company wins because it has a uniquely clever model. Salience wins on photonics hardware; Lumaril wins on access to plants and operators. In a world where foundation models are increasingly commoditized, the moat is the data, the deployment, and the relationship — the things that cannot be downloaded.
Expect investors to price physical risk differently. Deep tech demands longer timelines and larger checks, but commands premium valuations when the thesis lands, as Lumaril's round shows. Come prepared with real-deployment evidence — pilot results, throughput numbers, operator testimonials — not benchmarks and demos alone.
The bottom line
The AI funding market has not cooled; it has moved downstream from model training into the physical systems that make AI useful. Founders building where data meets machines, materials, and skilled labor are fishing where the biggest checks are written.
Angeled.com connects founders with investors.
Lumaril $66M Series A funding announcement at $400M valuation (led by Elad Gil; Felicis, Koch Disruptive, Lux, General Catalyst participating)
Prior coverage of Foundry Robotics $100M Series A and Bedrock Robotics $600M Series C
Read Next
General Translation Raises $27 Million Series A Led by a16z to Build Agent-First Localization Infrastructure
LangChain Raises $300 Million Led by CapitalG at a $3 Billion Valuation to Build the Agent Stack
Mount Thor Raises $45 Million to Run Cloud Fleets of Macs for AI Agents
OpenAI Reportedly Raises $15 Billion at a $1.25 Trillion Valuation
Q-CTRL Raises $290 Million Series C Led by Salesforce Ventures, Reaching Unicorn Status
SE3 Raises $30 Million Series A Led by Accel



