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

  • Zana za uhandisi wa programu – Bidhaa zinazozalisha kodi, kama vile Claude Code, mara nyingi hukwama kwa dakika kadhaa wakati modeli inachakata ombi. Kupunguza kusubiri huko hadi sekunde chache kutafanya maendeleo ya mwingiliano kuwa na mtiririko mzuri zaidi.
  • Mifumo ya usanifu wa kijenereta – Studio zinazobadilisha maelekezo ya maandishi kuwa mifano ya awali ya programu au rasilimali za michezo zinahitaji marudio ya haraka ili kuwafanya wabunifu waendelee kushiriki. Decoding ya haraka hupunguza mizunguko ya usanifu na kuongeza viwango vya ubadilishaji.

Sekta zote mbili zinageuza ucheleweshaji (latency) moja kwa moja kuwa upotevu wa saa za malipo au upotevu wa wateja, hivyo ongezeko la kasi mara 30 linaweza kuwa faida kubwa ya ushindani.

Mtazamo mpana zaidi

Mkakati wa Kog unapingana na mwelekeo wa tasnia wa kujenga silikoni maalum ya AI. Kampuni kama Cerebras huwekeza mabilioni katika chip zinazoahidi kiwango kikubwa cha utendaji kwa kila wati. Kog anahoji kuwa GPU za leo tayari zina upana wa mawasiliano ya kumbukumbu (memory bandwidth) wa kutosha kwa ajili ya decoding; sehemu inayokosekana ni programu inayoweza kuitumia kikamilifu. Ikiwa dai hili litathibitika kwa kiwango kikubwa, watengenezaji wanaweza kuahirisha maboresho ghali ya vifaa na kutegemea maboresho ya programu ili kufikia uchanganuzi (inference) wa karibu na wakati halisi.

Changamoto zinazoweza kujitokeza

  • Mzigo wa matengenezo – Kila kizazi kipya cha GPU kitahitaji uhandisi mpya wa kurejesha (reverse-engineering). Kadiri soko linavyotanuka, timu ndogo inaweza kulemewa.
  • Uwezo wa kutanuka kwa modeli kubwa zaidi – Onyesho la majaribio lilitumia modeli ya vigezo 2B. Kutanua mbinu zilezile kwenye mifumo mikubwa zaidi kunaweza kukumbana na mipaka ya uwezo wa kumbukumbu au kuhitaji mbinu za kihandisi za ziada ambazo bado hazijafichuliwa.
  • Njia mbadala – Watoa huduma za wingu tayari wanatoa mifumo iliyoboreshwa kwa ajili ya uchanganuzi (inference-optimized instances) inayounganisha chip maalum na programu. Kwa baadhi ya kazi, faida ndogo kutoka kwa mfumo wa Kog inaweza isiwe kubwa kuliko urahisi wa huduma inayodhibitiwa (managed service).

Mambo ya kuzingatia

  • Uanzishaji wa kwanza wa modeli kubwa – Kog anasema hatua yake inayofuata ni ongezeko la kasi mara 10 kwenye "modeli kubwa ya kiwango cha juu." Pengo kati ya onyesho la 2B na modeli ya kiwango cha biashara litakuwa mtihani wa wazi zaidi wa mbinu hiyo.
  • Upanuzi wa msaada wa vifaa – Kuongeza msaada kwa GPU mpya au viongeza kasi (accelerators) vingine kutakuwa ishara ikiwa modeli ya kiwango cha chini inaweza kwenda sambamba na mzunguko wa haraka wa vifaa.

Muhtasari

Kog inaonyesha kuwa kutoa utendaji zaidi kutoka kwa GPU zilizopo inawezekana unapoandika upya mfululizo wa programu (software stack) tangu mwanzo. Ahadi ya decoding ya LLM yenye kasi mara 30 zaidi inaweza kubadilisha jinsi watengenezaji wanavyoweka bei na kusanifu huduma za AI—mradi tu kampuni iweze kudumisha jitihada kubwa za kihandisi zinazohitajika kwa kila kipande kipya cha silikoni.