Mentium Technologies Inc.¶
Rad-hard neuromorphic AI coprocessor — UCSB spinout with deep NASA SBIR pipeline
Updated: Session 89 (2026-04-07)
Company Profile¶
| Field | Value |
|---|---|
| Location | Goleta, CA (Santa Barbara area) |
| Type | Semiconductor startup (UCSB spinout) |
| Founded | 2016–2017 |
| CEO | Mirko Prezioso (lead scientist for first integrated memristor neuromorphic hardware demo) |
| CTO | Farnood Merrikh Bayat |
| Co-founders | Dimitri B. Strukov, Konstantin Likharev |
| Employees | ~10-15 (estimated) |
| Funding raised | $1.05M (private) + $9.67M government |
| Product line | Luna-1 (2.4 TOPS, 70mW), Luna-R1 (rad-hard, 200kRad TID), Luna-2 (4.2 TOPS), Luna-R2 (rad-hard) |
Mentium was spun out of UCSB's Electrical and Computer Engineering department, specifically from the research group that published the first demonstration of neuromorphic computing hardware based on integrated memristors. The company builds ultra-low-power AI inference accelerators using in-memory computing with non-volatile memory, overcoming the data transfer bottleneck that limits conventional processors.
Key specs: 50 TOPS/W efficiency, <0.4W power consumption, radiation-hardened design.
Investors include: Alumni Ventures, Silicon Catalyst Ventures, Excellis, Terra.VC, and 7 others.
FO Project¶
155249 — Testing Neuromorphic Architectures for High Capacity/Low-Power AI in Sub Orbital Flight¶
- Status: Active (Apr 2023 – Apr 2026)
- TRL: 4 (began at 4, target 6)
- TX: TX02.1.3 High-Performance Processors
- ML Predicted TX: TX10.4.1 Verification and Validation of Autonomous Systems (mismatch — human classified as computing hardware, ML classified as autonomy application)
- Flight: Suborbital flight test
- Technology: Radiation-hardened AI inference accelerator coprocessor. Expands AI capabilities of existing space systems by orders of magnitude while consuming <0.4W. The FO project is the suborbital flight qualification of this chip.
TechPort Footprint (3 projects)¶
| Project | Program | Role | Period | TRL | Notes |
|---|---|---|---|---|---|
| 93616 | SBIR/STTR | PI (Prezioso) | Jun–Dec 2017 | 2→3 | Phase I — neuromorphic chip for UAS |
| 95729 | SBIR/STTR | PI (Merrikh Bayat) | Jul 2018 – Jun 2022 | 3→6 | Phase II — same, advanced to TRL 6 |
| 155249 | FO | PI (Prezioso) | Apr 2023 – Apr 2026 | 4→6 (target) | FO flight test — radiation-hardened version |
Lineage chain: Phase I [93616] → Advanced To → Phase II [95729] → Advanced To → FO [155249] (via Phase III SBIR). Classic SBIR pipeline with FO as the flight qualification step.
Funding¶
USASpending ($9.67M across 6 awards, all NASA)¶
| Award | Amount | Period | Description |
|---|---|---|---|
| 80NSSC22CA230 | $4.57M | Sep 2022 – Jan 2026 | SBIR Phase II Sequential — Neuromorphic chip for sensing, situational awareness, and decision making in radiative environments |
| 80NSSC26C0011 | $2.50M | Feb 2026 – May 2027 | SBIR Phase III — Enhancing and productizing the DVMR rad-hard digital AI coprocessor |
| 80NSSC23CA207 | $1.17M | Sep 2023 – Apr 2026 | SBIR Phase III — Testing neuromorphic architectures for high capacity/low-power AI in sub-orbital flight (= FO project funding) |
| 80NSSC18C0094 | $1.13M | Jul 2018 – Jun 2022 | SBIR Phase II — Low-power, ultra-fast deep learning neuromorphic chip for UAS |
| 80NSSC20C0682 | $181.0K | Sep 2020 – Mar 2022 | SBIR Phase I — Radiation hardened in-memory computing for space applications |
| NNX17CA38P | $116.7K | Jun–Dec 2017 | SBIR Phase I — Original neuromorphic chip concept |
Total NASA investment: $9.67M — substantial for a small semiconductor company.
Funding trajectory: $117K (2017) → $1.13M (2018) → $181K (2020, new rad-hard track) → $4.57M (2022, Phase II Sequential) → $1.17M (2023, Phase III/FO) → $2.50M (2026, Phase III productization). The 2026 Phase III ($2.5M for "enhancing and productizing the DVMR rad-hard digital AI coprocessor") is the commercialization milestone.
Private investment: $1.05M across 11 investors (Crunchbase/PitchBook)¶
No DoD contracts found in USASpending.¶
Technology Deep Dive¶
Architecture: Analog in-memory computing with non-volatile memory (memristors). This is fundamentally different from conventional digital processors — computation happens where data is stored, eliminating the memory-data transfer bottleneck ("von Neumann bottleneck") that limits conventional chips. The result is orders-of-magnitude better power efficiency for AI inference.
Radiation hardening: In June 2025, Mentium completed heavy ion radiation testing at the Berkeley Accelerator Space Effects (BASE) Facility's 88-inch Cyclotron, with NASA JPL and HPSC team consultation. Testing confirmed the processor can operate beyond LEO — validated for GEO and Lunar radiation environments. The Luna-R1 variant tolerates up to 200kRad TID.
Product lineup (as of 2026): | Model | TOPS | Power | Interface | Rad tolerance | Target | |-------|------|-------|-----------|---------------|--------| | Luna-1 | 2.4 | 70mW | GPIO | Commercial | IoT, edge AI | | Luna-R1 | 2.4 | 75mW | GPIO | 200kRad TID (RHBD) | LEO/GEO/Lunar | | Luna-2 | 4.2 | 130mW | xSPI, PCIe 3.0 | Commercial | Higher-performance edge | | Luna-R2 | 4.2 | 135mW | xSPI, PCIe 3.0 | 150kRad TID | Deep space missions |
All models supported by Mentium SDK — complete toolchain for compiling, quantizing, profiling, and deploying deep-learning models with high-fidelity on-chip emulator.
DVMR: The newest product designation — "Digital Voltage-Mode Resistive" — combines analog in-memory computing with a co-designed digital accelerator. This is the architecture being productized under the 2026 Phase III ($2.5M).
Synopsys partnership: Mentium uses Synopsys EDA tools for chip design (featured as a Synopsys success story).
Upstream Lineage¶
- Academic origin: UCSB ECE department — Strukov lab's integrated memristor research
- First demo: Prezioso led the first demonstration of neuromorphic hardware based on integrated memristors (published in Nature, 2015)
- UAS application: Original SBIR focused on drone autonomy (obstacle avoidance, target acquisition)
- Space pivot: Radiation hardening track started 2020 with separate SBIR Phase I, then merged with the main product line
Downstream — Active Deployment¶
Luna1 Selected for Odyssey Orbital Mission (March 2026)¶
SIGNIFICANT: NASA procured Mentium's Luna1 chip for deployment on the Odyssey mission — a 6U CubeSat launching in 2026. The chip will deliver real-time spaceborne analytics. This is Mentium's first orbital mission — a major step beyond the suborbital FO flight test.
NASA CCRPP Award (February 2026)¶
Mentium received a Civilian Commercialization Readiness Pilot Program (CCRPP) SBIR contract — the final stage of NASA's SBIR funding ladder, specifically designed to accelerate transition from R&D to commercial product. This is the strongest commercialization signal in the SBIR system.
DVMR Productization¶
The $2.5M Phase III (80NSSC26C0011, Feb 2026–May 2027) is explicitly for "enhancing and productizing" the rad-hard DVMR coprocessor. Combined with CCRPP, Mentium is in the active commercialization phase.
Other Applications¶
- Space: Onboard AI for sensor data analysis, autonomous control, system diagnosis — applicable to satellites, lunar systems, Mars missions
- Broader edge AI market: Defense, autonomous vehicles, IoT, robotics, security systems
- No DoD contracts yet — still surprising given the obvious defense applications, but CCRPP and Odyssey selection suggest commercial traction is building through NASA first
Assessment¶
| Dimension | Rating |
|---|---|
| Technology readiness | Advanced — TRL 6 on commercial version; rad-hard variant validated at BASE (200kRad TID); Luna1 selected for orbital mission |
| Funding trajectory | Strong — $9.67M NASA, $1.05M private, Phase III productization + CCRPP in 2026 |
| Team | Deep technical (memristor pioneers from UCSB) |
| Downstream impact | Active — Odyssey orbital deployment in 2026, CCRPP commercialization, full product lineup |
| Confidence | Confirmed — orbital mission selection + CCRPP award = active commercialization |
Time dimension: Founded 2016/17 → Phase I 2017 → Phase II 2018 → rad-hard pivot 2020 → Phase II Sequential $4.57M 2022 → FO flight test 2023 → BASE radiation testing June 2025 → Phase III productization $2.5M Feb 2026 → CCRPP award Feb 2026 → Odyssey orbital mission selection March 2026. Nine-year arc from founding to orbital deployment and active commercialization.
Key insight: Mentium has crossed the threshold from "promising SBIR company" to "active space product company." The Odyssey orbital mission is the inflection point — it's not a test, it's a NASA-procured deployment. Combined with the CCRPP award (NASA's strongest commercialization signal) and a four-model product lineup with full SDK, Mentium is no longer pre-commercial. The FO suborbital flight test was the qualification step; Odyssey is the operational step. Still no DoD contracts, but the space beachhead strategy appears to be working.
Timeline (updated)¶
| Year | Event |
|---|---|
| 2016–17 | Company founded (UCSB spinout) |
| 2017 | Phase I ($117K) — neuromorphic chip for UAS |
| 2018 | Phase II ($1.13M) — advanced to TRL 6 |
| 2020 | Rad-hard Phase I ($181K) — new track |
| 2022 | Phase II Sequential ($4.57M) — lunar/rad environments |
| 2023 | FO Phase III ($1.17M) — suborbital flight test begins |
| 2025 Jun | BASE radiation testing completed — JPL consultation, GEO/Lunar validated |
| 2026 Feb | Phase III productization ($2.5M) + CCRPP award |
| 2026 Mar | Luna1 selected for Odyssey orbital mission (6U CubeSat, launching 2026) |
| 2026 Apr | FO project completion (80NSSC23CA207) |
| 2027 May | DVMR productization Phase III completion |
Open Questions¶
- Has the FO suborbital flight actually occurred? (Project ends Apr 2026 — should be imminent or completed)
- What's the Odyssey mission profile — LEO, GEO, or beyond? What analytics will Luna1 perform?
- What's the DoD strategy — is Mentium pursuing SpaceWERX, DARPA, or AFRL contracts?
- What's the commercial pricing model — chip sales, IP licensing, or integration services?
- Has the $1.05M private funding been supplemented by new venture rounds?
Sources: TechPort 155249, 93616, 95729; USASpending; Mentium website; SBIR.gov; Crunchbase; Synopsys; NASA SBIR Newsletter
Session 89 update (2026-04-07): No changes since Session 60. TechPort [155249] still Active (period ends Apr 2026 — this month). TRL still shows 4 (target 6) — may not have been updated yet. USASpending unchanged ($2.5M Phase III + $1.17M FO Phase III as latest awards). No new contracts or public announcements found beyond Odyssey orbital mission (Feb/Mar 2026) and CCRPP (Feb 2026) already captured. FO project completing imminently — next refresh should confirm closeout and final TRL.