DeepSeek and Zhipu Reportedly Explore Custom AI Chips

Matthew Berman discusses 2026 reports of Chinese model developers pursuing their own processors and the strategic appeal of controlling more of their AI infrastructure.

MBMatthew Berman登壇者 · Forward Future
チャンネル
Matthew Berman
公開日
2026年7月15日
動画
0:42
読了時間
1 分
原文
英語
YouTube で開く
スクロールすると動画は右下に小さくなって再生を続けます。字幕は閲覧言語に合わせて切り替わります。
閲覧言語2 言語で読めます
この記事はまだ日本語に翻訳されていないため、英語の原文を表示しています。
要約

30 秒でわかる

Matthew Berman discusses reports that DeepSeek is developing an AI chip and that Zhipu is considering a custom processor. He sees reduced reliance on external suppliers as a strategic motivation. The reports describe development and exploration, not evidence that either company's proposed chip is already manufactured or deployed.

  1. 1A reported chip projectBerman says DeepSeek is reportedly working on a chip; the report does not establish a completed product.
  2. 2Reducing relianceHe argues that custom hardware could reduce dependence on Nvidia and give model developers more control.
  3. 3Zhipu also exploringBerman mentions reports about Zhipu, the company behind GLM-5.2. Model rankings depend on the evaluation.

ポイント

  1. Reports in July 2026 said DeepSeek was developing a custom AI chip and Zhipu was exploring one.
  2. Berman argues that proprietary chips could reduce dependence on external hardware suppliers.
  3. Designing a chip does not by itself guarantee control over fabrication, equipment or the whole supply chain.
  4. His description of GLM-5.2 as the best open model is an opinion, not a universal ranking.
全文

チャプターごとに整えた全文を読む

文字起こしのみ
01

DeepSeek and custom chips

Berman discusses reports of a DeepSeek chip project and its potential strategic appeal.

Matthew Berman discusses reports that DeepSeek, a Chinese AI model developer, is working on a custom AI chip. Reuters reported the development in July 2026 based on people familiar with the matter. The report concerns a project in development, not proof that the resulting chip has been manufactured or deployed.

Berman argues that making custom chips could reduce dependence on outside suppliers such as Nvidia and give a company more control over its AI stack. A chip design alone does not establish control over fabrication, equipment and the wider supply chain.

02

Another model developer

He says Zhipu is also reportedly considering an AI chip. The clip does not establish that either project has produced hardware.

He adds that Zhipu, the developer of GLM-5.2, is also reportedly considering its own AI chip. His description of the model as the best open model is his assessment, and the brief clip does not provide a benchmark establishing a universal ranking.

同じチャンネルの記事

スライド 12 枚34:21EN
Matthew Berman· 2026年10月27日Matthew Berman Examines Anthropic’s IPO Filing, AI Growth and GovernanceMatthew Berman, Sam Altman · Forward FutureThe presenter reviews the S-1’s financial metrics, compute strategy and voting structure while questioning the tension between rapid AI development and safety.AnthropicAI Safetyエンタープライズ AI+2
0:26EN
Matthew Berman· 2026年10月25日Sam Altman on Human Values and Problem-Solving in an AI WorldSam Altman · OpenAIIn a short interview excerpt, Altman considers the enjoyment of solving hard problems and the human role in interpreting results.Human-AI CollaborationFuture of WorkAI Strategy
スライド 2 枚0:36EN
Matthew Berman· 2026年10月25日Kimi K3.1 and a Rubik’s Cube Solver DemonstrationMatthew Berman · Forward FutureMatthew Berman shows a scrambled virtual cube being solved and gives the Kimi K3.1 test a favorable assessment.評価(Evals)コーディングエージェントOpen-weight models
スライド 2 枚0:21EN
Matthew Berman· 2026年10月24日An AI-Generated Kimi Benchmark Chart Used Invented ScoresMatthew Berman · Forward FutureMatthew Berman asks for a graph of Kimi K3.1 Pro benchmarks, then discovers that the displayed figures are illustrative placeholders.Benchmarks評価(Evals)Generative media
スライド 6 枚25:50EN
Matthew Berman· 2026年10月24日Kimi K3.1 Pro Tested: Strong 3D Demos, Uneven Results and Competitive PricingMatthew Berman · Forward FutureMatthew Berman tests the open-weight model on coding, 3D rendering and visual explanations, then compares benchmark results and prices.Open-weight modelsBenchmarksAI cost & performance+1
スライド 2 枚32:20EN
Matthew Berman· 2026年10月23日Who Is Local AI For? The Costs, Trade-offs and a Hybrid FutureMatthew Berman · Forward FutureMatthew Berman weighs data control and independence against hardware cost, setup complexity and inference speed.Local inferenceOpen-weight modelsAI cost & performance+1