UBS says 60% of companies now watching AI budgets are moving to cheaper models and open-source Chinese models
The pressure is coming from extreme bills, including users spending up to $35K/month, teams exceeding quotas by 200%, and companies cutting internal AI tools from 5 to 2.
Companies are not abandoning AI, they are using model routing, which sends easy tasks to cheaper models and saves premium models for hard reasoning, code, and long-context work.
Chinese open-source models such as Qwen, DeepSeek, MiniMax, GLM, and Kimi now fit the enterprise cost curve because they can be run locally or used through cloud catalogs.
The pressure is coming from extreme bills, including users spending up to $35K/month, teams exceeding quotas by 200%, and companies cutting internal AI tools from 5 to 2.
Companies are not abandoning AI, they are using model routing, which sends easy tasks to cheaper models and saves premium models for hard reasoning, code, and long-context work.
Chinese open-source models such as Qwen, DeepSeek, MiniMax, GLM, and Kimi now fit the enterprise cost curve because they can be run locally or used through cloud catalogs.
π1
π¨π Global One SI - Prompt
This is the most underrated Hermes agent prompt. Send this to Hermes, and your agent becomes your 24/7 personal executive assistant. Monitors your upcoming meetings, task list, travel updates & more. One of the best Hermes use cases
This media is not supported in your browser
VIEW IN TELEGRAM
Free Hermes guides, agent skills, memory & context tools, plugins, extensions, and so much more.
β hermesatlas.com
β hermesatlas.com
β1
This media is not supported in your browser
VIEW IN TELEGRAM
You are f*cking due for a big win bro keep your foot on the throttle Your time is coming
π¨π Global One SI - Prompt
Hardware & infrastructure price increases over the last year: RAM (DRAM) β Doubled since early 2025 β +90% in Q1 2026 alone β DDR5 sticks: $80 β $300 each GPU Memory (VRAM) β 16Gb module: $5.50 β $20+ β Now ~80% of a GPU's total cost GPUs β RTX 5090: $1β¦
If you want to buy a new computer and youβre waiting for prices to go down, you should probably stop waiting and just buy it now. Itβs gonna get worse before it gets better.
π₯1
π¨π Global One SI - Prompt
X-BAT's not just going where the wind takes it. The team is pushing X-BAT through thousands of wind tunnel test conditions, evaluating how it performs as the environment changes. The goal: a prototype engineered for stability, resilience, and mission successβ¦
This media is not supported in your browser
VIEW IN TELEGRAM
Shield AI has unveiled the X-BAT, a fully autonomous jet-powered aircraft that can take off and land vertically without a runway.
The aircraft can fly at altitudes of up to 50,000 feet and has a range of around 2,000 nautical miles.
Powered by Hivemind AI software, X-BAT is designed to carry out complex missions, operate in swarms, and work alongside piloted aircraft.
The aircraft can fly at altitudes of up to 50,000 feet and has a range of around 2,000 nautical miles.
Powered by Hivemind AI software, X-BAT is designed to carry out complex missions, operate in swarms, and work alongside piloted aircraft.
π¨π Global One SI - Prompt
So Apple now charges $199 (+54%) for a nearly 4 year old streaming TV box and worst of all blamed memory prices? This memory excuse only holds for its memory-heavy machines. A couple GB of RAM doesnβt cost $70 more on a TV. They repriced an old device dueβ¦
The Art of the deal
πΏ1
AI race is getting bigger..
CHINA UNVEILS ITS OWN AI RIVAL TO ANTHROPIC'S MYTHOS π
China's 360 security technology has introduced Yitian Tulong, a new AI cybersecurity platform designed as a domestic alternative to Anthropic's Mythos.
Company unveiled 2 powerful AI systems"
> Tulongfeng: An AI model that automatically discovers software vulnerabilities and has reportedly identified 3,432 vulnerabilities, including 105 verified by Chinese authorities.
> Yitianzhen: An AI-powered cyber defense system that helps detect threats, automate security operations, and respond to cyberattacks.
They're doing everything to stay in the race and even get ahead! The world is moving faster than ever!
CHINA UNVEILS ITS OWN AI RIVAL TO ANTHROPIC'S MYTHOS π
China's 360 security technology has introduced Yitian Tulong, a new AI cybersecurity platform designed as a domestic alternative to Anthropic's Mythos.
Company unveiled 2 powerful AI systems"
> Tulongfeng: An AI model that automatically discovers software vulnerabilities and has reportedly identified 3,432 vulnerabilities, including 105 verified by Chinese authorities.
> Yitianzhen: An AI-powered cyber defense system that helps detect threats, automate security operations, and respond to cyberattacks.
They're doing everything to stay in the race and even get ahead! The world is moving faster than ever!
π¨π Global One SI - Prompt
We've never seen anything like the product price increases announced today by Apple and Microsoft. Tim Cook: "I haven't seen anything like it in any area in over 40 years" Elon Musk agrees. The AI data center boom is driving a massive wave of consumer inflation.
Elon Musk just called the current AI chip price surge "the biggest price jump in anything I've ever seen."
Elon Musk said the production shortfall relative to demand is insane and that much higher production is needed.
This means the AI chip boom is not peaking yet. It is still in its early phase, with years of price pressure still ahead before supply catches up.
Elon Musk said the production shortfall relative to demand is insane and that much higher production is needed.
This means the AI chip boom is not peaking yet. It is still in its early phase, with years of price pressure still ahead before supply catches up.
π³1
π¨ AI MAY BE TRIGGERING THE THIRD WAVE OF INFLATION.
Even Tim Cook said this cost jump is something he hasn't seen in over 40 years in the business.
Tariffs can get negotiated. Oil prices can reverse when supply catches up. But this one doesn't work that way.
AI infrastructure spending isn't a price shock, it's a demand shock that's still in its early.
The five biggest hyperscalers are spending $741 billion this year, up 75% from last year, and most of that money hasn't even hit the ground yet.
That means the price pressure happening right now is just the opening phase, not the peak.
The mechanism is simple: AI needs memory chips. So does everything else.
Phones, consoles, cars, laptops all draw from the same limited chip supply, so when AI scales up its orders, it doesn't just raise AI costs, it raises the cost of nearly every electronic product on the market.
That's playing out in real time today.
Apple and Microsoft both announced price hikes on MacBooks, iPads, and Xbox consoles, the same day, citing the same memory shortage.
Nintendo and Sony already did the same thing weeks earlier.
This isn't an isolated company passing on costs, it's an entire hardware industry repricing around the same input shortage simultaneously.
The bet from the Fed is that AI eventually pays for itself through productivity gains that push inflation back down.
That bet might be right eventually. But UBS says that payoff is still years out, while the price increases are happening now.
Which means the Fed is being asked to hold rates steady or higher through a multi year stretch where the thing supposed to eventually fix inflation is actively making it worse in the meantime.
That's a much harder position for Warsh to defend than the temporary tariff and oil shocks the Fed has navigated so far.
Even Tim Cook said this cost jump is something he hasn't seen in over 40 years in the business.
Tariffs can get negotiated. Oil prices can reverse when supply catches up. But this one doesn't work that way.
AI infrastructure spending isn't a price shock, it's a demand shock that's still in its early.
The five biggest hyperscalers are spending $741 billion this year, up 75% from last year, and most of that money hasn't even hit the ground yet.
That means the price pressure happening right now is just the opening phase, not the peak.
The mechanism is simple: AI needs memory chips. So does everything else.
Phones, consoles, cars, laptops all draw from the same limited chip supply, so when AI scales up its orders, it doesn't just raise AI costs, it raises the cost of nearly every electronic product on the market.
That's playing out in real time today.
Apple and Microsoft both announced price hikes on MacBooks, iPads, and Xbox consoles, the same day, citing the same memory shortage.
Nintendo and Sony already did the same thing weeks earlier.
This isn't an isolated company passing on costs, it's an entire hardware industry repricing around the same input shortage simultaneously.
The bet from the Fed is that AI eventually pays for itself through productivity gains that push inflation back down.
That bet might be right eventually. But UBS says that payoff is still years out, while the price increases are happening now.
Which means the Fed is being asked to hold rates steady or higher through a multi year stretch where the thing supposed to eventually fix inflation is actively making it worse in the meantime.
That's a much harder position for Warsh to defend than the temporary tariff and oil shocks the Fed has navigated so far.
β1
π¨π Global One SI - Prompt
Let this actually sink in: The most powerful closed models on earth now require federal sign-off, one user at a time. AI might've just hit its first real wall for progression. Meanwhile, open-source models (like GLM-5.2 and Sakana Fugu): β Can't beβ¦
Hereβs a fun comparison between GLM 5.2 and Opus 4.8 on a one-shot reproduction of the SDPO paper
This is a hard task: the model must resolve messy verl issues and then run ablations to completion and confirm the paperβs claims.
- GLM 5.2 costs $6.21 while Opus 4.8 cost us $46.35
- Both models spent a bulk of their tokens resolving initial verl issues. GLM 5.2 attempted 14 failed runs before first success while Opus 4.8 attempted 9 runs.
- GLM 5.2 surprisingly took 2.65M tokens (excl re-reads) compared to 4.53M tokens for Opus 4.8
This is a hard task: the model must resolve messy verl issues and then run ablations to completion and confirm the paperβs claims.
- GLM 5.2 costs $6.21 while Opus 4.8 cost us $46.35
- Both models spent a bulk of their tokens resolving initial verl issues. GLM 5.2 attempted 14 failed runs before first success while Opus 4.8 attempted 9 runs.
- GLM 5.2 surprisingly took 2.65M tokens (excl re-reads) compared to 4.53M tokens for Opus 4.8
π¨βπ»1
The future of AI is undoubtedly open-sourced.
For 99% of people, the top open-weight models can already handle all your daily needs.
Imo, the best thing you can do right now is run a "barbell" approach for your AI usage:
Intial 10% and planning phase: Use frontier intelligence (Opus/GPT)
Gruntwork and the middle 80%: Switch to open-weight model (GLM, Kimi)
Final 10% and review: Switch back to froniter intelligence to verify
Best of both worlds - cost efficiency + intelligence.
For 99% of people, the top open-weight models can already handle all your daily needs.
Imo, the best thing you can do right now is run a "barbell" approach for your AI usage:
Intial 10% and planning phase: Use frontier intelligence (Opus/GPT)
Gruntwork and the middle 80%: Switch to open-weight model (GLM, Kimi)
Final 10% and review: Switch back to froniter intelligence to verify
Best of both worlds - cost efficiency + intelligence.
β1
This media is not supported in your browser
VIEW IN TELEGRAM
How it feels to argue in your head while having two different mindsets
China Robotics Startup Map 2026
β€ China Robotics:
β’ 2000+ robotics startups
β 50+ humanoid robot companies
β $13B+ funding from 2015 to 2025
β£ Humanoid robots
β Unitree
β AgiBot
β Fourier
β EngineAI
β£ Industrial robots
β Estun
β Efort
β Siasun
β£ Service robots
β Pudu
β Keenon
β OrionStar
β£ Core components
β Inovance
β Hanβs Robot
β LEADOR
β£ AI and robot software
β Megvii
β RealAI
β DEEP Robotics
β£ Main hubs
β Beijing
β Shanghai
β Shenzhen
β Suzhou
β Hangzhou
β Dongguan
The important part is not only the humanoid body.
China is building the full robotics stack around it.
Hardware.
AI software.
Components.
Factories.
Supply chain.
Capital.
Pilot deployments.
That is where the real robotics race is moving.
β€ China Robotics:
β’ 2000+ robotics startups
β 50+ humanoid robot companies
β $13B+ funding from 2015 to 2025
β£ Humanoid robots
β Unitree
β AgiBot
β Fourier
β EngineAI
β£ Industrial robots
β Estun
β Efort
β Siasun
β£ Service robots
β Pudu
β Keenon
β OrionStar
β£ Core components
β Inovance
β Hanβs Robot
β LEADOR
β£ AI and robot software
β Megvii
β RealAI
β DEEP Robotics
β£ Main hubs
β Beijing
β Shanghai
β Shenzhen
β Suzhou
β Hangzhou
β Dongguan
The important part is not only the humanoid body.
China is building the full robotics stack around it.
Hardware.
AI software.
Components.
Factories.
Supply chain.
Capital.
Pilot deployments.
That is where the real robotics race is moving.
This media is not supported in your browser
VIEW IN TELEGRAM
Pov: The failure of the family arrives at the family dinner
Volkswagen is preparing to cut 15% of its entire global workforce.
VW is now weighing up to 100,000 job cuts on top of a global headcount of 667,164, and shutting four German factories, Hanover, Zwickau, Emden, and Audi's Neckarsulm.
This is the biggest restructuring in the company's 88-year history.
The supervisory board discusses this on July 9. And this is the third escalation in 18 months.
Late 2024: 35,000 jobs.
March 2026: raised to 50,000 after operating profit collapsed 53% to β¬8.9 billion, the worst result since the 2016 Dieselgate scandal.
Now: 100,000, double that number, with plant closures added on top for the first time in the company's history.
Despite that 44% profit collapse in 2025, VW's executive board still received bonus payouts totaling β¬13.6 million, with CEO Oliver Blume personally taking home β¬7.4 million.
The bonuses were tied to cash flow, not profit, so management got paid while the company posted its worst year since Dieselgate.
China is the main reason this is happening.
VW was China's top automaker for years. BYD took that spot in 2024. VW fell to third place in 2025, behind Geely.
Non-Chinese automakers' share of the Chinese EV market collapsed from 57% in 2020 to 32% in 2025. VW's own EV sales inside China fell 44.3% in a single year.
Chininese makers aren't staying in China only. BYD, Chery, SAIC, and Leapmotor doubled their combined European market share through May, expanding directly into VW's home turf in Europe while VW is losing ground in China at the same time.
VW plans to cut planned investment 15% to just over β¬130 billion over five years, while reportedly considering splitting the core VW brand and parts business into separate entities.
Unions are already promising war.
IG Metall and VW's works council said Friday they'll "do everything in our power to prevent" the cuts, the same union that organized 100,000 workers in warning strikes across Germany in December 2024, shutting down production at multiple plants for hours at a time, over a round of cuts that was half this size.
If that's what 35,000 cuts triggered, a fight over 100,000 jobs and four shuttered factories is likely to be far more disruptive.
VW is now weighing up to 100,000 job cuts on top of a global headcount of 667,164, and shutting four German factories, Hanover, Zwickau, Emden, and Audi's Neckarsulm.
This is the biggest restructuring in the company's 88-year history.
The supervisory board discusses this on July 9. And this is the third escalation in 18 months.
Late 2024: 35,000 jobs.
March 2026: raised to 50,000 after operating profit collapsed 53% to β¬8.9 billion, the worst result since the 2016 Dieselgate scandal.
Now: 100,000, double that number, with plant closures added on top for the first time in the company's history.
Despite that 44% profit collapse in 2025, VW's executive board still received bonus payouts totaling β¬13.6 million, with CEO Oliver Blume personally taking home β¬7.4 million.
The bonuses were tied to cash flow, not profit, so management got paid while the company posted its worst year since Dieselgate.
China is the main reason this is happening.
VW was China's top automaker for years. BYD took that spot in 2024. VW fell to third place in 2025, behind Geely.
Non-Chinese automakers' share of the Chinese EV market collapsed from 57% in 2020 to 32% in 2025. VW's own EV sales inside China fell 44.3% in a single year.
Chininese makers aren't staying in China only. BYD, Chery, SAIC, and Leapmotor doubled their combined European market share through May, expanding directly into VW's home turf in Europe while VW is losing ground in China at the same time.
VW plans to cut planned investment 15% to just over β¬130 billion over five years, while reportedly considering splitting the core VW brand and parts business into separate entities.
Unions are already promising war.
IG Metall and VW's works council said Friday they'll "do everything in our power to prevent" the cuts, the same union that organized 100,000 workers in warning strikes across Germany in December 2024, shutting down production at multiple plants for hours at a time, over a round of cuts that was half this size.
If that's what 35,000 cuts triggered, a fight over 100,000 jobs and four shuttered factories is likely to be far more disruptive.
π₯1