While Cerebras builds custom AI chips, French startup Kog squeezes performance out of the GPUs enterprises already own. By rewriting low-level software for existing datacenter hardware, Kog eliminates the latency bottlenecks that slow AI workflows.

Challenging the Necessity of Custom AI Silicon

The AI industry has pivoted to purpose-built chips for inference. Kog's CEO, Gaël Delalleau, says the belief that standard GPUs can’t handle decoding is wrong. Modern GPUs—Nvidia H200, AMD MI300X—offer memory bandwidth that current software leaves idle.

Kog’s Kog Inference Engine (KIE) targets "extremely fast single-request decoding," a must-have for real-time apps. In a recent demo the company hit 3,000 tokens per second (TPS) with Laneformer 2B, an open-source 2-billion-parameter model tuned for their stack. The demo uses a small model, but Kog plans to extend those gains to much larger LLMs.

A "Hacker" Approach to GPU Engineering

Unlike hardware-agnostic providers such as ZML, Kog digs deep. The team reverse-engineers GPUs at the assembly and binary levels, treating the silicon as a set of physical laws to master.

That low-level focus extracts every ounce of efficiency, but it costs time. With a lean crew of 11 engineers, Kog spends weeks or months dissecting each new GPU architecture before adding support. The method delivers top-tier performance but limits how quickly Kog can cover new hardware.

Targeting High-Value AI Use Cases

Kog focuses on sectors where latency equals lost revenue:

  • Software engineering: Tools like Claude Code stall for minutes. Kog aims to turn "waiting hours" into near-instant results.
  • Generative app/game design: Prompt-to-app pipelines need rapid iteration to keep users engaged and revenue flowing.

The company’s next milestone is a 10× speedup on its first major large-scale model. Backed by Scaleway, Bpifrance and the French Tech 2030 program, Kog positions itself as a key player in Europe’s AI sovereignty drive.

Key Takeaways

  • Optimization over hardware: Kog pushes the memory bandwidth of Nvidia H200 and AMD MI300X GPUs to the limit with extreme low-level software engineering.
  • Breaking the latency barrier: The startup targets a 30× boost in LLM inference speed for real-time professional workflows such as automated coding and generative design.
  • Deep-level engineering: By adopting a "hacker" mindset, Kog reverse-engineers GPU assembly and binary code to achieve performance that standard stacks cannot reach.

Kog, a French startup, says its Kog Inference Engine aims to run large-language-model (LLM) decoding up to 30 times faster on the same Nvidia H200 or AMD MI300X GPUs that data-center operators already own.

Why the push for speed matters now

LLM inference now throttles products like code-completion assistants, on-the-fly content generators, and interactive game-design tools. In those settings, the gap between a user’s prompt and the model’s reply directly impacts productivity and revenue. High-level APIs such as CUDA leave a large slice of GPU memory bandwidth idle, especially during token-by-token decoding, which dominates real-time use cases.

Kog’s low-level answer

Kog skips conventional software layers and talks straight to the silicon. Its engineers reverse-engineer the GPU at the assembly level, treating the hardware as a set of constraints to obey rather than a black box to abstract. The result is a custom execution path that keeps data flowing through the GPU’s memory pipes at near-full capacity.

In a recent demo the company ran Laneformer 2B—a 2-billion-parameter open-source model tuned for its stack—and reported a high token throughput. The model is modest compared with larger commercial systems, but the speed gain shows what a purpose-built software stack can extract from the same hardware.

The engineering trade-off

The upside comes with a steep cost curve. Kog’s 11-engineer team spends weeks or months dissecting each new GPU architecture before it can be supported. That depth yields high performance on the targeted cards, but it also means Kog cannot instantly add support for every new GPU that hits the market. Customers benefit only if their fleets include the Nvidia H200 or AMD MI300X GPUs Kog has already optimized.

Who stands to gain

  • Software-engineering tools – Products that generate code, such as Claude Code, often stall for minutes while the model processes a request. Cutting that wait to a few seconds would make interactive development far more fluid.
  • Generative design pipelines – Studios turning textual prompts into app prototypes or game assets need rapid iteration to keep creators engaged. Faster decoding shortens design loops and boosts conversion rates.

Both sectors translate latency directly into lost billable hours or churn, so a 30× speed boost could be a decisive competitive edge.

The broader picture

Kog’s strategy runs counter to the industry trend of building custom AI silicon. Companies like Cerebras pour billions into chips that promise higher throughput per watt. Kog argues that today’s GPUs already have enough memory bandwidth for decoding; the missing piece is software that can actually use it. If the claim holds at scale, developers could defer costly hardware upgrades and rely on a software upgrade to achieve near-real-time inference.

Potential headwinds

  • Maintenance burden – Every new GPU generation will require fresh reverse-engineering. As the market diversifies, the small team could be stretched thin.
  • Scalability to larger models – The demo used a 2 B-parameter model. Scaling the same techniques to much larger systems may hit memory-capacity limits or demand additional engineering tricks not yet disclosed.
  • Alternative paths – Cloud providers already offer inference-optimized instances that bundle specialized chips and software. For some workloads, the marginal gain from Kog’s stack may not outweigh the convenience of a managed service.

What to watch

  • First large-model rollout – Kog says its next milestone is a 10× speedup on a “major large-scale model.” The gap between the 2 B demo and an enterprise-grade model will be the clearest test of the approach.
  • Hardware support expansion – Adding support for newer GPUs or other accelerators will signal whether the low-level model can keep pace with rapid hardware cadence.

Takeaway

Kog shows that squeezing more work from existing GPUs is possible when you rewrite the software stack from the ground up. The promise of 30× faster LLM decoding could reshape how developers price and architect AI services—provided the company can sustain the intense engineering effort needed for each new piece of silicon.