OpenAI puts Token Cost Pressure Front and Center for the First Time. The next round of AI competition shifts from showcasing capabilities to focusing on usage costs.
1｜Altman acknowledges Token Cost is now a 'Huge Issue'
At a corporate event, Sam Altman admitted that AI token cost has become a 'huge issue'. According to Axios, OpenAI's largest token user consumes about 1 trillion tokens per month. What used to be a growth myth last year is now a cost alarm today. Enterprise clients are no longer asking how smart the model is, but rather how much budget a single agent workflow can burn through.
While OpenAI has been emphasizing over the past year that 'AI will only get cheaper', the agentification has turned this promise on its head. Chatbots are priced per use, while agents burn tokens along task chains. Behind what seems like an automated process could lie continuous calls, retrievals, code execution, and multi-round validations. Cost pressure is no longer just a model company's gross margin issue; it will determine whether companies dare to integrate AI into core workflows. The first hurdle to AI commercialization is not whether users will adopt it, but whether the bill can be sustained long-term.
(Source: Axios / Tom's Hardware)
2｜Apple's New Siri Version Takes Lessons from Google and NVIDIA; TSMC says Advanced Chips Won't Be Enough for Years to Come
Apple's AI delay is turning into a supply chain problem. According to The Information's headline and summary, the new Siri version scheduled for September will rely on Google's NVIDIA chip cluster. The signal of this arrangement is not in Siri itself, but in Apple relying on an external computing network for the first time to power its consumer-facing AI revival. Apple excels at locking in critical capabilities into proprietary systems, but AI is pushing it back to cloud-based supply chains.
On the same day, TSMC CEO C.C. Wei stated after a shareholders' meeting that demand for AI-driven advanced chips will continue to outpace supply for the next few years. Even with new U.S. factories coming online, they will not be able to completely satisfy U.S. customer demand. AMD's Helios MI455X platform has been exposed, but early systems still use UALink-over-Ethernet, potentially dragging down performance. The AI competition appears to be about model iteration on the surface, but at its core lies the queuing rights of advanced processes, HBM, interconnects, and packaging.
(Source: The Information / The Verge / Reuters / Tom's Hardware)
3｜In May, US Tech Industry Saw 38,242 Layoffs, While Anthropic Claims 80% of New Code Generated by Claude
In May, the US tech industry experienced approximately 38,242 layoffs, reaching a nearly two-year monthly high, with AI being one of the most commonly cited reasons in corporate announcements. When viewed alongside figures for tech giants' capital expenditures, the sting is even more pronounced. Companies have not ceased investing in technology; rather, they have shifted their budgets from personnel to models, chips, and data centers. The labor market is making room for AI capital expenditure.
Anthropic's latest figures make this trend even more explicit. The company stated that in May, over 80% of the new code integrated into production was generated by Claude, with engineers' quarterly code delivery volume increasing eightfold compared to the 2021-2025 baseline. This metric does not mean that programmers have been replaced, but it has altered management's conception of productivity. Once "fewer people enabling more agents" becomes boardroom language, layoffs are no longer just a cyclical adjustment but a restructuring of organizational dynamics.
(Source: Tom's Hardware / VentureBeat / Synced)
4｜Benchmark Establishes First Growth Fund, Ramp Raises $44 Billion at $440 Billion Valuation, AI Turns VC Discipline into Late-Stage FOMO
Benchmark has raised a $2 billion new fund, with $1.25 billion of it being its inaugural growth fund. The significance of this move lies in Benchmark's reputation over the past two decades for its roughly $425 million flagship fund and early-stage discipline. Now, it is also preparing ammunition for mature startups. AI has prompted the most restrained type of Silicon Valley capital to accept larger funds and later-stage risks.
On the same day, Ramp raised $7.5 billion at a $44 billion valuation. It is not a pure-play AI model company but one that embeds AI into corporate finance operations, expense management, risk control, and payment flows. Capital is favoring these types of companies because they are closer to budget entry points than general-purpose chatbots. Benchmark's late-stage fund and Ramp's valuation together illustrate that AI investment is shifting from "who has a model" to "who can take over enterprise workflows." Valuations are no longer just buying technology; they are increasingly buying distribution and account positioning.
(Source: TechCrunch / Axios / The Information)
5｜Bitcoin Falls Below $62,000 as AI Stocks and Gold Chip Away at the Liquidity Narrative
Bitcoin dropped below $62,000 on June 4, with about $1.5 billion in crypto longs liquidated within 24 hours. Citing CoinGlass data, CoinDesk reported that over 208,000 traders were liquidated, with BTC-related losses exceeding $800 million and ETH-related losses around $386 million. This wasn't a single bad news-triggered flash crash but more like a deleveraging event meeting liquidity reprice.
Presto Research's explanation was straightforward. Investors are reallocating funds to gold and AI stocks while reassessing the Fed rate cut outlook. Yesterday's stablecoin payment platform announcement indicated that crypto infrastructure is still being absorbed by traditional finance, but today's price signal suggests that BTC's macro narrative hasn't automatically trumped AI and gold. As AI becomes the stock market's main theme, Bitcoin's "digital gold" identity needs to reassert itself.
(Source: CoinDesk / CoinGlass / Presto Research)
Also worth noting ↓
AI Companies Call on US Congress to Enhance Bioweapon Defense. Executives from Anthropic, OpenAI, Microsoft, and other companies signed an open letter urging measures to address AI-assisted bioweapon risks. Model companies are shifting security issues from self-regulation statements to legislative agendas. (Source: The Verge)
SpaceX Receives 35-Year Full Property Tax Exemption for $50 Billion Terafab Semiconductor Factory in Texas. The approved tax incentive's value may reach hundreds of millions of dollars, with residents pushing back against the cost transfer of AI and aerospace manufacturing to local fiscal politics. (Source: Tom's Hardware)
Waymo to Repurpose Retired Robotaxi Batteries for Grid Energy Storage. The self-driving car fleet is transitioning from traffic assets to energy assets, embedding battery lifecycles into the grid narrative. (Source: TechCrunch)
Trump Announces $800 Million Support Plan for Coal Industry. AI data centers are driving up power demand, prompting coal to reenter the policy toolbox in the name of energy security and grid stability. (Source: Washington Post)
Cash App Introduces NFC Payment Wand. Payment hardware is shifting from efficiency tools to social accessories as fintech continues to package transaction actions into consumer symbols. (Source: TechCrunch / The Verge)
Axios reports that Trump's second term is expanding the scope of U.S. overseas military operations. The tension between the "America First" mantra and the actual military footprint is growing, and congressional checks and balances pressure may continue to increase. (Source: Axios)
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
