Page 3 of 7 · 153 stories in this section
ByteDance, Nvidia's largest Chinese customer, is ramping AI infrastructure investments. Meanwhile, MiniMax's M3 model achieves 15.6x faster decoding at 1M tokens, and Armada raises $131M for modular edge data centers.
ByteDance is offering special stock incentives to its AI team to combat poaching, while becoming Nvidia's largest Chinese customer with early, large bets on AI infrastructure.
ByteDance is granting special stock to its AI team to retain talent as rivals poach staff. The company has become Nvidia's largest Chinese customer after early bets on AI infrastructure.
Researchers removed safety guardrails from Meta and Google AI models in minutes, highlighting fragility. Meanwhile, JVP exited DealHub at over 6x return.
Xunce launched TokenONE, the first TokenOS, enabling AI models to directly invoke enterprise data as tokens. The vertical-domain token pricing ranges from $10 to $100 per million tokens, aiming to boost gross margins fro
Xunce launched TokenONE, the first TokenOS, on May 25, 2026, pricing data tokens at $10–100 per million—over ten times the cost of general-purpose AI models.
Xunce (3317.HK) launched TokenONE, the world's first TokenOS, on May 25, 2026. It enables enterprise data to be directly invoked by AI models as data Tokens, targeting a 20-30% reduction in data governance costs by end o
Xunce launched TokenONE, the first TokenOS, enabling enterprise data as AI tokens. Priced at $10-100 per million tokens, it aims to lift gross margins from 10-20% to 20-30% by end of 2026.
Xunce (3317.HK) launched TokenONE on May 25, 2026, a TokenOS that turns dormant enterprise data into AI fuel, targeting a reduction in data preparation time from 60-80% to 20-30% by end of 2026.
Wiwynn showcases rack-scale AI servers with CPO optical interconnects and liquid cooling at Computex 2026, integrating Ayar Labs' TeraPHY engines and SuperNova light sources.
DHL Supply Chain switched to an agentic architecture after standard LLMs failed at numerical reasoning for warehouse design. The move succeeded where earlier models would have been too slow, highlighting the need for gov
AI agents deployed in enterprises introduce new risks, with every action potentially causing infrastructure chaos. DHL's early warehouse optimization attempts struggled with numerical reasoning, highlighting the need for
WEA Japan, a Web3 payment infrastructure firm, is integrating AI-driven risk control into its stablecoin platform for real-time settlement. The company holds over 30 core patents in secure payments and encrypted communic
Lenovo's AI-related revenue soared 84% in Q4, driving shares up nearly 20%. Meanwhile, Chinese AI startup DeepSeek attracts investment from CATL in its first funding round.
Corsair unveiled the CORSAIR PRO portfolio of AI workstations and servers, including the FlexPrime V80B with NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, targeting model development and deployment.
Alibaba's Qwen Team releases Qwen3.7-Max, a proprietary model capable of 35 hours of continuous autonomous execution, marking a strategic retreat from open source as pricing wars with OpenAI and Anthropic intensify.
SpaceX's AI infrastructure unit is valued at $165 billion, with Anthropic paying $15 billion annually for data center access. CORSAIR also enters AI hardware with its new PRO portfolio.
Major pharmaceutical companies are accelerating AI investments, with deals ranging from Pfizer-PostEra's $610M partnership to Nvidia's $1B tie-up with Eli Lilly, as the industry bets AI can slash drug development timelines.
Anthropic agreed to pay SpaceX $1.25 billion per month through May 2029 for cloud compute, totaling $15 billion annually. Meanwhile, Alibaba released proprietary Qwen3.7-Max, capable of 35 hours of autonomous execution.
Cheaper AI models from Chinese labs like DeepSeek and open-source alternatives are pressuring OpenAI and Anthropic, whose IPOs are valued over $800 billion each. Meanwhile, pharma giants like Pfizer and Eli Lilly ink bil
SpaceX's IPO filing shows a $15 billion annual AI compute deal with Anthropic and governance ties to Elon Musk loyalists. Meanwhile, Imperagen raises £5 million for quantum-AI enzyme engineering.
Google unveiled Gemini 3.5 Flash at I/O, targeting $1 billion in annual savings and processing 1 trillion tokens per day. The model is designed to be smart, fast, and inexpensive, shifting 80% of workloads.
Chinese labs like DeepSeek match frontier AI at lower cost, pressuring OpenAI and Anthropic's $800B+ IPO valuations. Western challengers including Cohere and Nvidia counter with cheaper, trusted alternatives.
Anthropic is in early discussions to rent Azure servers powered by Microsoft's custom Maia 200 chips, deepening ties between the AI startup and Microsoft as its OpenAI partnership evolves.