Key Takeaways
- Chennai/Singapore-based fabless semiconductor startup optoML has closed a $1.8 million pre-Series A round co-led by deep-tech investors Bluehill.VC and a99.
- The company has completed a 12 nm tapeout with TSMC, the final design step before chip fabrication, and signed an MoU with Kaynes Semiconductor for chip assembly and testing.
- optoML’s patented analog-in-memory compute architecture promises up to 50x higher energy efficiency versus conventional digital AI accelerators.
- The capital will fund team expansion and the development of next-generation AI System-on-Chip (SoC) platforms targeting edge devices, enterprise infrastructure, and hyperscale data centres.
Quick Recap
Deep-tech semiconductor startup optoML has officially raised $1.8 million in a pre-Series A funding round, co-led by Bluehill.VC and a99. The announcement, shared by NewnexHQ on X (formerly Twitter), confirms the round will accelerate hiring and fund the next generation of AI-focused chip development. The funding comes shortly after optoML completed a critical 12 nm tapeout with TSMC, a milestone that validates its chip design and moves the startup closer to commercial production.
optoML Secures $1.8 Million in Pre-Series A to Propel Next-Gen AI Chip Innovationhttps://t.co/nHeLCIVb2P#optoML, #PreSeriesA, #Funding, #Semiconductors, #AIChips, #SoC, #DeepTech, #Innovation, #Singapore, #Bengaluru, #AIEfficiency, #ChipDevelopment, #BluehillVC
— NEWNEX – VC Syndication & Co-Investment Platform (@NewnexHQ) February 25, 2026
optoML’s Analog-in-Memory Architecture
optoML, founded in 2024 by Saravana Maruthamuthu, is not building another conventional digital AI accelerator. The company is developing analog-in-memory computing architectures integrated with optical interconnects, a fundamentally different approach to processing AI workloads. In traditional chip designs, data must travel back and forth between memory and processing units, creating latency and wasting energy.
optoML’s architecture performs computations directly within memory and uses optical interconnects on-chip to move data with minimal power loss. Maruthamuthu brings over 17 years of semiconductor and systems experience spanning Intel, Qualcomm, Continental, and Isar Aerospace. The startup operates under a fabless model, meaning it designs the chips while outsourcing manufacturing. Its scalable AI SoC platforms are built on advanced FinFET process nodes.
optoML claims its patented technology can deliver up to 50x higher energy efficiency compared to traditional digital accelerators. Manu Iyer, General Partner at Bluehill.VC, stated that optoML’s 12 nm tapeout with TSMC and its partnership with Kaynes Semicon “mark a critical transition from research to real silicon and scalable production”. The MoU with Kaynes Semiconductor, which recently rolled out India’s first commercially manufactured multi-chip module from its OSAT facility in Gujarat, provides optoML with a clear path to chip assembly and testing once wafers arrive from TSMC.
Why This Funding Is Timely?
The timing of this raise aligns with a pressing reality in the AI industry: energy consumption is becoming a first-order constraint. As AI models grow larger and deployments expand from cloud data centres to edge devices, the demand for chips that deliver high performance per watt has never been greater. Traditional GPU-based architectures, while powerful, consume enormous energy. This has pushed investors and researchers alike toward alternative compute paradigms such as analog-in-memory, photonic, and neuromorphic processing.
India’s semiconductor ecosystem is also maturing at exactly the right moment. Kaynes Semiconductor’s OSAT facility in Sanand, Gujarat, which began mass production in 2025, gives startups like optoML domestic access to chip assembly and testing infrastructure that previously required going entirely offshore. Government initiatives under the India Semiconductor Mission are providing tailwinds for deep-tech ventures.
Vignesh Shankar, General Partner at a99, noted: “As AI workloads move to the edge, power efficiency becomes the real constraint. optoML’s analog-in-memory approach is designed to address exactly that”. For hyperscalers and edge deployers alike, a 50x improvement in energy efficiency, if validated at scale, would represent a significant shift in the cost structure of AI inference.
Competitive Landscape
optoML operates in a growing field of startups pursuing alternatives to conventional digital AI chips. Two of the most relevant competitors at a comparable stage are TetraMem (analog in-memory computing using memristor/RRAM technology) and Opticore (photonic computing using optical processing units). Below is a comparison across key metrics.
| Feature / Metric | optoML | TetraMem | Opticore |
| Core Technology | Analog-in-memory compute + optical interconnects | Analog in-memory computing using multi-level RRAM/memristors | Photonic computing using optical processing units (OPUs) |
| Energy Efficiency Claim | Up to 50x vs. digital accelerators | Up to 100x TOPS/W vs. leading AI chips | Up to 100x vs. CMOS electronics |
| Process Node | 12 nm FinFET (TSMC tapeout completed) | 22 nm SoC | Standard foundry processes (node not publicly specified) |
| Target Applications | Edge, enterprise, data centres | Edge AI, IoT, smart audio, industrial systems | Data centres, hyperscale AI |
| Total Funding | ~$1.8M (pre-Series A) | Undisclosed (partnered with SK hynix) | ~$14.5M (seed + extension) |
| Manufacturing Model | Fabless; TSMC fabrication, Kaynes assembly | Fabless; commercial foundry CMOS integration | Fabless; standard foundry processes |
| Key Partnership | Kaynes Semicon (assembly & testing) | SK hynix (joint research on in-memory computing) | Backed by Origin Ventures, Jetha Global |
Strategic Analysis
optoML occupies a unique position by combining both analog-in-memory compute and optical interconnects in a single architecture, while TetraMem focuses purely on memristor-based analog compute and Opticore focuses purely on photonic processing.
TetraMem and Opticore hold an edge in stated efficiency claims (100x versus optoML’s 50x), but optoML’s completed 12 nm TSMC tapeout and its Kaynes Semicon partnership give it a clearer near-term path to production-grade silicon, especially within the India manufacturing ecosystem.
Bayelsa Watch’s Takeaway
I think this $1.8 million raise is a meaningful signal, not for its size, but for what it validates. In my experience covering deep-tech semiconductor deals, the real story is rarely the dollar figure at the pre-Series A stage. It is whether the technology has moved past the whiteboard. optoML’s completed 12 nm tapeout with TSMC tells me this is no longer a research project. It is a chip that is heading toward real silicon.
I find the combination of analog-in-memory compute with optical interconnects particularly interesting. Most competitors in this space pick one lane. optoML is attempting to merge two, which is ambitious but also risky. If the 50x energy efficiency claim holds up in real-world deployments, I believe this startup could attract significantly larger follow-on rounds, especially as hyperscalers and edge AI companies hunt for alternatives to power-hungry digital accelerators.
Sources
