Twitch Turns On Amazon AI Training by Default: 'Nobody Would Opt In'
Related
Rare Books Traced to Amazon AI Training Facility to Be Scanned and Destroyed
A tracking device planted in a book order ended at a Las Vegas facility where Amazon strips bindings to scan pages for training data.
Amazon's hardware chief: The company will accelerate the development of its own edge AI chips to prepare for device chip upgrades.
According to Mars Finance, Amazon's chief hardware executive recently revealed that the company is focusing on developing its own chips for key consumer devices such as the Echo Show and Fire TV. Last October, Amazon launched the AZ3 and AZ3 Pro chips, designed to enable AI models to run locally on devices. For Amazon, this focus on custom chips is also part of its broader strategy to improve the performance of on-device AI. (Cailian Press)
Ming-Chi Kuo: Amazon's consumer electronics products are shifting towards self-developed processors to optimize cost structure.
Analyst Ming-Chi Kuo stated that Amazon plans to develop its own processors for consumer electronics products to prepare for the ongoing wave of AI infrastructure development, aiming to reduce chip costs and allocate sufficient funds for AI investments. Amazon's free cash flow was only around $1.2 billion in Q1 2026. Starting in 2027, Amazon plans to gradually adopt a COT (Customer Own Tool) model, similar to its AI server chip Trainium, to develop the processors it needs, with Worldchip responsible for backend design and testing. It is projected that after the full transition, annual shipments could reach 40 million units. (Sina Finance)
The FTC has approved Musk's antitrust filing for Mesh's acquisition, which involves the deployment of optical networks for AI data centers.
According to ChainCatcher, the latest filings with the U.S. Federal Trade Commission (FTC) indicate that Elon Musk has received antitrust approval to acquire optical networking startup Mesh Optical Technologies. This means the FTC has completed a expedited antitrust review and will not question the transaction on competition grounds, clearing a major regulatory hurdle for the deal. However, it has not yet been disclosed whether the deal has been signed or completed. Mesh was founded by former SpaceX engineers, and its core product is optical transceivers for AI data centers. Compared to traditional network hardware, these transceivers improve energy efficiency, reduce latency, and enhance reliability to meet the demand for millions of optical connections driven by the growth of AI computing clusters. The founding team previously participated in the development of the laser communication system for SpaceX's Starlink satellite network and plans to deploy optical communication technology in space in the future to meet the inter-satellite laser communication needs of orbital data centers and AI satellite networks. The company completed a funding round of over $50 million in February of this year, led by Thrive Capital. The acquisition of Mesh is one of SpaceX's moves to strengthen its competitiveness in large-scale computing clusters. SpaceX has now made AI computing power a core business segment. Its xAI division operates the Colossus and Colossus II training clusters, totaling approximately 1GW of computing power, making it the first company to deploy a continuous gigawatt-scale AI training cluster. Colossus II will add over 400MW of computing power and more than 220,000 GB300 chips. It has already signed computing power cooperation agreements with Anthropic, Google, and Reflection AI, directly competing with hyperscale cloud providers such as Amazon Web Services, Microsoft Azure, and Google Cloud. This year, SpaceX also reached a Terafab chip manufacturing agreement with Tesla and Intel, extending its vertical integration capabilities in chip design and manufacturing. Over the past week, SpaceX's stock price ended its upward trend, closing at $153.23 per share, a drop of over 32% from its peak of $225.64 per share.
Coinbase has cut AI spending by nearly half and is trying to make open weight models such as GLM 5.2 and Kimi 2.7 the default option.
According to Mars Finance, on June 27th, Coinbase CEO Brian Armstrong stated that the key to maintaining stable AI spending while token usage grows exponentially lies not in setting usage friction or spending alerts, but in better default models, routing, and caching mechanisms. Coinbase is experimenting with defaulting to open-weight models such as GLM 5.2 and Kimi 2.7 through its LLM gateway, while still encouraging engineers to choose the appropriate model based on the task. He stated that 91% of employees never reach their usage limits, so instead of lowering limits and increasing alerts, the company has shifted to lower-cost default models. Regarding model routing, Coinbase preprocesses prompts in custom processes and routes tasks to the most suitable model based on cache hit rate and model pricing. For example, a cutting-edge model might be needed during the planning phase, but overuse during the execution phase could be excessive. He believes that in the future, models should not be chosen by humans; AI can automatically complete this task. Armstrong also stated that cache misses are the easiest way to drive up costs. Coinbase's requests are cache-aware to reuse hot caches as much as possible. For example, after correctly implementing caching, LibreChat's cache hit rate has improved from 5% to 60%. Furthermore, Coinbase requires engineers to keep contexts concise, including opening new sessions when switching tasks, narrowing file context scope, and disconnecting unused tools. The goal is not to suppress AI usage, but to build infrastructure that can support exponential growth. Through these practices, Coinbase has reduced its AI spending by nearly half, while token usage continues to grow.
Serenity analyzes the AI strategies of tech giants: optimistic about Amazon's development path, while Microsoft and Meta need to prove the necessity of their capital investments.
According to Odaily Odaily, Serenity, the "white-haired stock guru," stated in an article on the X platform that the market should not interpret the AI capital expenditures of large technology companies as "funds being withdrawn." A more accurate description of these investments is that they are for achieving large-scale revenue growth or profit margin improvement in the future. He currently favors Amazon the most, believing it to be one of the clearest AI transformation cases among hyperscale cloud service providers. He believes that Amazon may reduce operating costs in the future by using large language models to achieve autonomous delivery, warehouse robots, and logistics and transportation automation. At the same time, Amazon is also driving revenue growth by expanding AWS computing power and may enter the AI chip sales market with its self-developed chip, Trainium. Serenity believes Google ranks second among tech giants in AI strategy, with its AI capital investment aimed at protecting its search business's competitive advantage. Meanwhile, Google Cloud, leveraging its TPU computing power, also possesses the commercial potential of chips similar to Nvidia's GPUs. Regarding Microsoft and Meta, Serenity states that both companies still need to demonstrate the necessity of their large-scale AI capital investments to the market. Microsoft's recent lagging progress on its self-developed AI chip, Maia, and the impact of its AI development pace due to its collaboration with OpenAI have led to weaker market sentiment.