DeepSeek V4 Launches on Huawei Chips; Cognition AI Targets $25B Valuation

DeepSeek’s pivot to Huawei hardware and Cognition’s $25 billion funding bid signal two inflection points in the AI arms race — one geopolitical, one commercial.

🤖 DeepSeek V4 Launches on Huawei Chips, Shifting China’s AI Hardware Allegiance

Decoded: Chinese AI startup DeepSeek released a preview of its V4 model on April 24 — its first flagship since R1 disrupted global AI markets in January 2025. The V4 was built in close collaboration with Huawei and runs natively on Huawei’s Ascend AI accelerators, a direct departure from DeepSeek’s prior reliance on Nvidia GPUs. The V4 Pro version outperforms all public open-source models on world-knowledge benchmarks, trailing only Google’s closed-source Gemini Pro 3.1. The model is optimized for AI agent tasks requiring long, complex reasoning chains and became the fastest-trending model on Hugging Face upon release. Nvidia CEO Jensen Huang warned this month: “The day that DeepSeek comes out on Huawei first, that is a horrible outcome for our nation.” An Omdia semiconductor analyst called the collaboration “a big deal,” noting Huawei’s Ascend chips are China’s best domestic alternative to Nvidia. (Reuters, April 24, 2026)

Why it matters: DeepSeek’s pivot to Huawei Ascend hardware is the clearest signal yet that China’s AI stack is decoupling from Nvidia’s ecosystem. U.S. export controls since 2022 restricted Nvidia’s most capable chips from Chinese customers — but DeepSeek’s R1 showed competitive models could be trained cost-efficiently on constrained hardware. V4 extends that trajectory: if China’s leading AI labs can deliver frontier-competitive models on domestic silicon, export restrictions become a progressively weaker moat. For Nvidia (NVDA), China represented over 13% of FY2024 data center revenue before export tightening; further displacement from the Chinese AI stack shrinks both current revenue and the long-term developer ecosystem that underpins Nvidia’s platform advantage.

💰 AI Coding Agent Cognition Seeks $25 Billion Valuation in New Funding Round

Decoded: Cognition AI, maker of the Devin autonomous software engineering agent, is in early talks to raise a new funding round at a $25 billion valuation — more than doubling its previous $2 billion mark — Bloomberg reported April 23. No terms have been finalized. Cognition launched in 2024 with Devin, marketed as the first AI software engineer capable of independently completing multi-step coding tasks: planning, writing, testing, and deploying code without human intervention. The company competes directly with GitHub Copilot (Microsoft), Cursor, and Claude Code (Anthropic). The $25 billion figure reflects sustained investor appetite for autonomous coding agents as software development automation shifts from autocomplete to full task execution. (Bloomberg, April 23, 2026)

Why it matters: A $25 billion valuation for a two-year-old startup with a single product would make Cognition one of the highest-valued private AI companies globally. The figure reflects how aggressively the market is repricing autonomous software development: as agentic coding tools move from autocomplete to full task completion, incumbents with large developer bases — GitHub, JetBrains, Cursor — face direct displacement risk. For public market investors, the private comp signals continued pressure on Microsoft’s GitHub Copilot business and validates Anthropic’s Claude Code as a strategic asset. It also anchors what acquirers — Google, Meta, Amazon — might pay to control the autonomous coding category before it consolidates.

Stay decoded. See you tomorrow.

— The Get AI Decoded Team