Tech Express2026-08-070 次浏览0 条评论

Tech Express | August 7, 2026 – Tesla & SpaceX Announce $16.8B Terafab Chip Factory in Texas

语音朗读
--:--

1. Tesla & SpaceX Officially Announce $16.8B 'Terafab' Chip Factory in Texas

Tesla and SpaceX have formally announced their jointly-developed advanced chip manufacturing facility, Terafab, will be located in Grimes County, Texas, just north of Houston. The initial phase carries a massive $16.8 billion investment, with Elon Musk declaring the goal is to "bring cutting-edge manufacturing back to America" and build one of the world's largest chip fabrication plants.

The factory will focus on producing next-generation AI training and inference chips for Tesla's autonomous driving systems and SpaceX's Starship program. Following months of speculation, the formal announcement sent ripples through markets and is widely seen as a landmark moment in the push to reshore semiconductor manufacturing to the United States. Future expansion phases could push the total investment significantly higher.

Tesla & SpaceX Officially Announce Terafab

2. OpenAI Makes GPT-5.6 Free for All Users with Unlimited Text Chats

OpenAI announced major upgrades on August 6: GPT-5.6 Luna will become the default model for all free ChatGPT users, with the removal of text chat limits entirely. Paid users of GPT-5.6 Sol also received enhanced reasoning capabilities, including a new "thinking slider" that lets users control how deeply the AI deliberates on each response.

This represents OpenAI's most aggressive free-tier expansion to date, meaning non-paying users can now access near-frontier-level AI conversation capabilities. Analysts noted that the AI industry's competition is rapidly shifting from paywalls to ecosystem capture, with Google's Gemini and Anthropic's Claude already offering powerful free versions.

3. OpenAI, Microsoft, Amazon & Others Unite on Agent Plugins Standard

On August 6, OpenAI, Microsoft, Amazon, Cursor, and Vercel jointly launched Agent Plugins 1.0.0 — an open, vendor-neutral standard for packaging AI agent extensions. Under the new specification, developers build a plugin once and it runs across ChatGPT, Copilot, Cursor, and other compatible platforms.

The move marks a pivotal shift from the AI model arms race to ecosystem standardization. The format has been described as "USB for the AI era," dramatically reducing developer adaptation costs and accelerating the AI agent application ecosystem. Google and AWS have already committed to supporting the standard.

Agent Plugins Standard

4. ByteDance Begins Training 10-Trillion-Parameter AI Model

According to the Financial Times citing three sources, ByteDance has officially launched a next-generation large model pre-training project codenamed "Unknown," targeting up to 10 trillion parameters. If completed at this scale, the model would be over three times larger than China's current largest model (Moonshot AI's Kimi K3).

The project is in its early stages and represents a critical step for Chinese AI companies in the super-scale model race. ByteDance's Doubao and Jimeng products have already amassed large user bases domestically and internationally. Industry estimates suggest training a 10-trillion-parameter model could cost over $1 billion.

ByteDance 10-Trillion-Parameter Model

5. Memory "Big Three" Sell Out 2027 Capacity as AI Chip Demand Surges

Multiple reports confirm that Samsung, SK Hynix, and Micron have fully allocated their entire DRAM and HBM production capacity for 2027, with NAND Flash capacity also nearly sold out. Industry insiders say some AI companies are already "begging for chips," and 2027 could be the "most severe" year for memory shortages.

Nvidia is reportedly considering reducing the memory capacity of its next-generation Rubin Ultra chip to cope with the HBM shortage. The insatiable demand for high-performance memory has pushed major manufacturers to sign 3-to-5-year long-term supply agreements with key customers, transforming the memory market from a cyclical commodity into a seller-dominated scarce resource.

Memory Big Three Sell Out 2027 Capacity

6. Trevor Noah to Host Google Pixel 11 Launch on August 12

Google has confirmed its annual Made by Google hardware event will take place on August 12 in New York, hosted by comedian Trevor Noah. The star-studded guest lineup includes NBA superstar Stephen Curry, podcast queen Alex Cooper, Indian cricket star Shubman Gill, and Japanese actress Ayami Nakajo.

The Pixel 11 series will debut the "HiLight" active illumination feature and support 55% battery charge in 5 minutes. The Pixel 11 Pro is expected to feature an enhanced camera system and new AI capabilities, setting the stage for an intensifying premium smartphone battle with Apple's upcoming iPhone 18.

Trevor Noah Pixel 11 Launch

7. OpenAI's First Smart Speaker: Jony Ive-Designed, $300-$400 Price Tag

According to Bloomberg's Mark Gurman, OpenAI's upcoming first AI hardware device is a hockey puck-sized, doughnut-shaped smart speaker designed by io, the startup helmed by former Apple design chief Jony Ive. The device has no display, fits in one hand, features "moving parts that help give it personality," and is priced between $300 and $400.

This is significantly more expensive than mainstream smart speakers from Amazon and Google ($40-$240), but analysts believe that if it delivers AI interaction far beyond existing products, it could pioneer a new premium AI hardware category.

OpenAI Smart Speaker

8. U.S. Scientists Use AI to Design Complete Viral Genomes for the First Time

Researchers from Stanford University have announced the first successful AI-driven design of complete viral genomes. Using AI models Evo 1 and Evo 2, the team generated 302 designs and selected 16 that successfully replicated in the lab — bacteriophages engineered to kill E. coli bacteria without posing any threat to humans.

The findings, published in the journal Science, mark a milestone at the intersection of synthetic biology and AI. However, multiple experts have issued urgent warnings that the same technology could be weaponized to create novel diseases, calling for an international governance framework for AI-designed biological systems.

AI-Designed Viruses

9. China's First CXL 3.1 and PCIe 6.0 SSD Controllers Debut Simultaneously

At the Flash Memory Summit (FMS) 2026, Chinese storage controller maker Innogrit simultaneously unveiled the country's first CXL 3.1 controller and first PCIe 6.0 SSD controller, both having completed tape-out in the first half of 2026. This marks a breakthrough for domestic storage controllers in both cutting-edge interconnect standards.

The CXL 3.1 controller will provide higher-bandwidth, lower-latency memory expansion for AI data centers, while the PCIe 6.0 controller pushes SSD read/write speeds to new heights. Amid the global AI compute boom, advancements in domestic controller chips are critical for securing China's AI infrastructure supply chain.

China CXL PCIe Controllers

10. ByteDance's Volcano Engine Launches Seedance 2.5 Video Model API

On August 7, ByteDance's Volcano Engine officially launched the API for Doubao Seedance 2.5, its latest video generation model. Compared to its predecessor, version 2.5 delivers dramatic improvements in instruction following, long-form narrative coherence, photorealism, and audiovisual quality, natively supporting 30-second video generation with a maximum of 50 seconds.

ByteDance continues to push aggressively in the AI video generation space, with the Seedance series becoming a benchmark product in China's domestic market. As large language models become commoditized through free tiers, multimodal AI — particularly video generation — is becoming the core battlefield for differentiation among AI companies.

Seedance 2.5

更多优惠

🏠最新电商折扣

评论

评论 (0)

0/500
暂无评论