Home / Directory / AI infrastructure / Runtimes, compilers & systems software / TeRAM
TeRAM
Los Altos semiconductor company designing three-dimensional SRAM that sits with an AI compute chip, so inference can read model data with less waiting on memory.
Chip designers and inference-hardware teams planning for memory bandwidth and power, who can wait until the end of the decade for first production.
Anyone ordering inference servers now, or a software team looking for a runtime they can call this year.
Verdict
TeRAM is a Los Altos semiconductor company designing custom three-dimensional SRAM that can be integrated directly with AI compute chips. The bet is that inference reads enormous amounts of model data over and over, so memory bandwidth and power decide a large share of the cost. The buyers it is aimed at are the companies that will build those chips and the systems around them. First customer production is targeted for 2029, so this is a part on a multi-year clock, not a component in a server order placed now.
TechStartups reported on September 15, 2026 that TeRAM emerged from stealth with $37 million in equity financing, which Primary Venture Partners describes as a seed round. The round was co-led by Primary Venture Partners, B Capital, Hyperion, and SemiAnalysis Capital, with Alumni Ventures and Lightscape Partners participating. The financing closed on September 10 and was announced on September 15. Total disclosed funding is $37 million (TechStartups).
Score Breakdown
How TeRAM scores in the categories that matter to its buyers.
Pricing
TeRAM is a seed-stage chip project with first customer production targeted for 2029, so there is no part a buyer can order today (TechStartups).
The Field at a Glance
Where TeRAM ranks among Runtimes, compilers & systems software vendors we reviewed, by Overall Score and relative typical engagement cost.
TeRAM scores 6.3 in this peer set, below d-Matrix (6.7), Moreh (7.5), and Groq (7.9). Relative cost sits with a semiconductor development program, not with an API bill.
Use-case matrix
| Use case | Fit | Notes |
|---|---|---|
| On-package memory for an inference chip | Strong | 3D SRAM integrated directly with the compute chip. |
| Cutting the cost of reading model weights repeatedly | Strong | That repeated read is the stated reason for the design. |
| A part in a 2026 or 2027 server build | Weak | Customer production is targeted for 2029. |
| A software runtime or model host | Weak | TeRAM is a memory device, not a service you call. |
| Manufacturing proof | Mixed | The round funds a design that still has fabrication work ahead. |
Review Excerpts
Representative buyer views drawn from public information about TeRAM. Paid reviews and pay-for-play sites were excluded.
“Putting 3D SRAM on the same package as the compute chip targets the cost of reading model weights again and again.”
Representative buyer view
“An inference-chip team would compare the memory path with HBM for weight-heavy models.”
Representative buyer view
“The company is aiming at customer production in 2029, so a buyer ordering inference hardware now cannot use it.”
Representative buyer view
Where customer reviews couldn't be found, our analysts attempted to give a voice to buyer sentiment, drawing on real information from review sites, forums, and product documentation.
Based on: Tech Startups’ funding roundup
Methodology
This page is an independent evaluation of TeRAM for buyers comparing options in runtimes, compilers & systems software. AI Industry Reviews accepts no sponsorships, advertising, or pay-for-placement fees. TeRAM did not pay for this review.
What we scored
The headline number is an Overall Score on a 0-10 scale. Eight criteria fall under it in two groups.
- Buyer outcomes (for runtimes, compilers & systems software)
- Memory architecture for inference
- Fit to repeated model reads
- Evidence the part can be manufactured
- Time until a customer can buy it
- Company & commercial
- Innovation & product leadership
- Project management & communication
- Pricing
- Contract fairness
Pricing measures whether the price looks fair for the value delivered, including packaging and renewal friction that show up in real buying cycles.
Score Composition
| Input | Weight | What it covers |
|---|---|---|
| Reviews | 40% | A proprietary read of what practitioners say about likes, complaints, and day-to-day use, including public review sites, forums, and private chat rooms. Paid reviews and pay-for-play sites are out of scope. |
| Product | 35% | Hands-on look at screens and workflows. |
| Pricing | 15% | Whether the price looks fair for what you get. |
| Docs & training | 10% | Docs, tutorials, and training material. |
How we balanced the evidence
The Overall Score is the simple average of the eight criteria. Recommendation language follows that score and the use-case matrix above.
Scope
TeRAM is graded here as runtimes, compilers & systems software. Criteria scores can move as more review volume and product checks are added.