# Taalas hardwired LLM chip

45 Hacker News commenters recommended Taalas hardwired LLM chip and 8 criticized it, November 2025 to November 2026. It was recommended in 8 of 13 months. It meets the Hacker Approved badge rule.

**Verdict:** People recommend it for very fast LLM inference from a model etched into silicon. Critics note the chip is locked to one model and updates are limited.

## Why people recommend it

- Extremely fast token generation speed (5 comments): "Their demo is almost unbelievably fast, but as I understand it, the limitation of Taalas's strategy is KV-cache." ([source](https://news.ycombinator.com/item?id=48672419)) "implemented Llama3 8b on hardware to achieve 17000 tokens/s at a small fraction of the usual power consumption" ([source](https://news.ycombinator.com/item?id=48380365)) "Trying Taalas is almost scary, there is something unsettling with that speed!" ([source](https://news.ycombinator.com/item?id=48520525))
- Lower power use than GPUs (4 comments): "When other companies, like Taalas are putting research dollars in to making AI scalable, affordable and efficient." ([source](https://news.ycombinator.com/item?id=48693872)) "implemented Llama3 8b on hardware to achieve 17000 tokens/s at a small fraction of the usual power consumption" ([source](https://news.ycombinator.com/item?id=48380365)) "Those taalus chips apparently run at 1/10 the power as the current SOTA GPU setups." ([source](https://news.ycombinator.com/item?id=48673280))
- Strong proof of concept (3 comments): "proving out that baked silicon models can be truly instant" ([source](https://news.ycombinator.com/item?id=47240056)) "Taalas - conceptually, anyway - is on to something." ([source](https://news.ycombinator.com/item?id=47086927)) "Taalas has a pretty impressive proof of concept" ([source](https://news.ycombinator.com/item?id=48874376))
- Useful for agents and moderation (3 comments): "It seems like this would be useful right now for some things (eg moderation)." ([source](https://news.ycombinator.com/item?id=49443503)) "at 17k tokens per second it's likely enough to reliably call tools" ([source](https://news.ycombinator.com/item?id=48447835)) "In a coding agent it is a meaningful improvement." ([source](https://news.ycombinator.com/item?id=48663969))

## What people warn about

- Locked to one model, hard to update (4 comments): "the ability for it to be updated to a brand-new model as new models come out is limited" ([source](https://news.ycombinator.com/item?id=48663864)) "Taalas is a nice concept, but I don’t want to use the same model forever!" ([source](https://news.ycombinator.com/item?id=47986920)) "Taalas process is more expensive And not really worth it because once you burn a model on the silicon, the silicon can only serve that model." ([source](https://news.ycombinator.com/item?id=47147462))
- Power efficiency claims unproven (3 comments): "Low latency is nice. But it would be more interesting if they could demonstrate the efficiency of energy consumption." ([source](https://news.ycombinator.com/item?id=48668261)) "the worry with that suggestion is that it would blow out energy per net unit of work" ([source](https://news.ycombinator.com/item?id=47819062)) "It doesn't add up with their claims of better power efficiency." ([source](https://news.ycombinator.com/item?id=47087028))
- Weak on large or current models (4 comments): "taalas is great for llama 3.x 8B models, really bad for one board serving Kimi K3" ([source](https://news.ycombinator.com/item?id=49205337)) "the Taalas product is not a fair comparison because it runs an old outdated model" ([source](https://news.ycombinator.com/item?id=48693280)) "taalas is super fast but that model is so dumb" ([source](https://news.ycombinator.com/item?id=49557912))
- Costly, unclear economics (2 comments): "So this is very cool. Though I'm not sure how the economics work out?" ([source](https://news.ycombinator.com/item?id=47087440)) "Taalas process is more expensive And not really worth it because once you burn a model on the silicon, the silicon can only serve that model." ([source](https://news.ycombinator.com/item?id=47147462))

## Alternatives hackers recommend

- [Google TPU](https://hackerapproved.com/p/google-tpu.md): 73 recommend
- [Taalas HC1](https://hackerapproved.com/p/taalas-chip.md): 59 recommend
- [Cerebras wafer-scale chip](https://hackerapproved.com/p/cerebras-wafer-scale-chip.md): 29 recommend
- [Cerebras CS-4](https://hackerapproved.com/p/cerebras-cs4.md): 12 recommend
- [Groq LPU](https://hackerapproved.com/p/groq-lpu.md): 9 recommend

## Where to get it

- [Visit the official site](https://taalas.com/)

## Cite as

Hacker Approved, "Taalas hardwired LLM chip", https://hackerapproved.com/p/taalas/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).