The AI capex cycle is now pushing memory into a much more aggressive pricing upcycle than expected.
UBS now expects DRAM prices to rise 32% quarter-over-quarter in Q3 and 18% in Q4, while NAND prices are expected to rise 30% in Q3 and 12% in Q4.
AI infrastructure is eating memory supply faster than the industry can add it. High Bandwidth Memory, DRAM, and NAND are all getting pulled into the same demand wave. UBS also sees the DRAM market staying undersupplied until at least 2028, because demand growth in 2027 is expected to be far above supply growth.
UBS now expects DRAM prices to rise 32% quarter-over-quarter in Q3 and 18% in Q4, while NAND prices are expected to rise 30% in Q3 and 12% in Q4.
AI infrastructure is eating memory supply faster than the industry can add it. High Bandwidth Memory, DRAM, and NAND are all getting pulled into the same demand wave. UBS also sees the DRAM market staying undersupplied until at least 2028, because demand growth in 2027 is expected to be far above supply growth.
Sonnet 5 might be the worst-value model on this chart...
Cost per intelligence index task:
>1.3x more expensive than opus 4.8 max
>2.2x more expensive than gpt-5.5 xhigh
>4.8x more expensive than glm-5.2 max
>7.4x more expensive than kimi k2.6
>57x more expensive than deepseek v4 pro
And it still isn’t the strongest model here...
You’re not paying for better intelligence.
You’re paying the anthropic tax.
Cost per intelligence index task:
>1.3x more expensive than opus 4.8 max
>2.2x more expensive than gpt-5.5 xhigh
>4.8x more expensive than glm-5.2 max
>7.4x more expensive than kimi k2.6
>57x more expensive than deepseek v4 pro
And it still isn’t the strongest model here...
You’re not paying for better intelligence.
You’re paying the anthropic tax.
✍1
🇨🇭 Global One SI - Prompt
POV: How bros with $17.38 in their portfolio feel when Trump announces Tarrifs
Everyone said tariffs would kill US manufacturing.
The data says otherwise.
US ISM Manufacturing PMI (50 = expansion line):
2022-2025:
→ Mostly stuck below 50 for 2.5 straight years
→ Brief pokes above 50, never sustained
2026:
→ Broke above 50 in February
→ 5 straight months above the line
→ Hit 54 in June, the highest print since mid-2022
The predicted collapse never showed up in the data.
Europe should be reading this chart, not the headlines.
The data says otherwise.
US ISM Manufacturing PMI (50 = expansion line):
2022-2025:
→ Mostly stuck below 50 for 2.5 straight years
→ Brief pokes above 50, never sustained
2026:
→ Broke above 50 in February
→ 5 straight months above the line
→ Hit 54 in June, the highest print since mid-2022
The predicted collapse never showed up in the data.
Europe should be reading this chart, not the headlines.
🇨🇭 Global One SI - Prompt
How to turn Hermes into your personal assistant (save this)
Claude Fable is insane for UI/UX design work.
Inject this prompt into Claude Code, and you'll be able to ship Apple-style liquid glass design into ANY vibe-coded workflow:
(works in apps, websites, dashboard & more)
Inject this prompt into Claude Code, and you'll be able to ship Apple-style liquid glass design into ANY vibe-coded workflow:
(works in apps, websites, dashboard & more)
✍1
You cannot build an AI data center without gas power.
Even renewable-first governments admit it now.
Global new gas power capacity jumped 31% in a year, to 1,047 GW in 2025:
🇺🇸 US: 252 GW in development, a third tied to data centers, 2024 gas restrictions abolished in 2025
🇨🇳 China: 153 GW in development
🇩🇪 Germany: passed a 9 GW gas bidding law in May, reversing years of gas reduction
🇯🇵 Japan: more long-term LNG procurement for AI and chip demand
🇻🇳 Vietnam: named LNG a core pillar of its 2030 power plan
🇰🇷 South Korea: Samsung's VC publicly requesting more LNG for new fabs
Renewables didn't replace gas and AI made gas necessary again.
Even renewable-first governments admit it now.
Global new gas power capacity jumped 31% in a year, to 1,047 GW in 2025:
🇺🇸 US: 252 GW in development, a third tied to data centers, 2024 gas restrictions abolished in 2025
🇨🇳 China: 153 GW in development
🇩🇪 Germany: passed a 9 GW gas bidding law in May, reversing years of gas reduction
🇯🇵 Japan: more long-term LNG procurement for AI and chip demand
🇻🇳 Vietnam: named LNG a core pillar of its 2030 power plan
🇰🇷 South Korea: Samsung's VC publicly requesting more LNG for new fabs
Renewables didn't replace gas and AI made gas necessary again.
“Claude code will kill software engineers”...
Absolutely not.
The best software engineers will earn 50-100x more than average developers...
Because their job is shifting from writing every line of code to leading teams of ai agents.
>Design the architecture
>Coordinate multiple agents
>Review their output
>Catch failures before production
>Turn generated code into reliable systems
The more agents you can control and manage effectively, the more valuable you become.
Ai makes the gap between average and exceptional engineers much bigger.
Absolutely not.
The best software engineers will earn 50-100x more than average developers...
Because their job is shifting from writing every line of code to leading teams of ai agents.
>Design the architecture
>Coordinate multiple agents
>Review their output
>Catch failures before production
>Turn generated code into reliable systems
The more agents you can control and manage effectively, the more valuable you become.
Ai makes the gap between average and exceptional engineers much bigger.
✍1
This media is not supported in your browser
VIEW IN TELEGRAM
When you spend 4 years studying hard for an IT degree only to become a nurse because AI took your job before graduation
🇨🇭 Global One SI - Prompt
Silicon valley: ai is the biggest money machine ever Meanwhile... Estimated cumulative ai spend vs revenue: 1. Amazon: $331b spent → $22b revenue 2. Google: $303b spent → $25b revenue 3. Meta: $238b spent → $3b revenue 4. Microsoft: $256b spent → $31b revenue…
The AI circular bet
Forwarded from 🛢WTI, Brent Crude Oil, NGAS Trading 📊
🌍 One map explains why geopolitics moves markets.
It shows each country’s number 1 export.
🔹Oil in the Middle East and Russia.
🔹Electronics and machinery in China.
🔹Capital goods, aerospace and telecom in the US.
🔹Cars and parts across Germany, Europe and Canada.
🔹Copper in Chile. Coal in Australia. Crops, minerals and energy across Africa and South America.
The point is simple: globalization is not evenly spread.
It depends on a few critical nodes: energy, factories, metals, logistics and high-value industrial tools.
Disrupt one of them with tariffs, sanctions, war, shipping delays or export controls and the shock does not stay local.
It becomes inflation, commodity volatility and a global repricing of risk.
The next winners will not just be countries with resources.
They will be the ones that control processing, logistics, technology and supply-chain redundancy.
It shows each country’s number 1 export.
🔹Oil in the Middle East and Russia.
🔹Electronics and machinery in China.
🔹Capital goods, aerospace and telecom in the US.
🔹Cars and parts across Germany, Europe and Canada.
🔹Copper in Chile. Coal in Australia. Crops, minerals and energy across Africa and South America.
The point is simple: globalization is not evenly spread.
It depends on a few critical nodes: energy, factories, metals, logistics and high-value industrial tools.
Disrupt one of them with tariffs, sanctions, war, shipping delays or export controls and the shock does not stay local.
It becomes inflation, commodity volatility and a global repricing of risk.
The next winners will not just be countries with resources.
They will be the ones that control processing, logistics, technology and supply-chain redundancy.
🇨🇭 Global One SI - Prompt
Claude Fable is insane for UI/UX design work. Inject this prompt into Claude Code, and you'll be able to ship Apple-style liquid glass design into ANY vibe-coded workflow: (works in apps, websites, dashboard & more)
How to build a self-evolving AI brain with Hermes agent x Obsidian
✍1
A list of useful AI workflows to tackle this weekend (pick a couple):
• Hermes agent setup
• Second-brain via Claude Code x Obsidian
• TradingView MCP connection (for traders)
• Deep financial research via perplexity finance
• Personal OS dashboard w/ Fable 5
• Learn loop engineering (full article on this page)
• NotebookLM for research *any topic*
• Set up Claude Skills for your repetitive workflows
• Build Notion database as your central hub (then insert agents to manage it)
SO much to do in AI right now.
• Hermes agent setup
• Second-brain via Claude Code x Obsidian
• TradingView MCP connection (for traders)
• Deep financial research via perplexity finance
• Personal OS dashboard w/ Fable 5
• Learn loop engineering (full article on this page)
• NotebookLM for research *any topic*
• Set up Claude Skills for your repetitive workflows
• Build Notion database as your central hub (then insert agents to manage it)
SO much to do in AI right now.
✍1
🇨🇭 Global One SI - Prompt
How to build a self-evolving AI brain with Hermes agent x Obsidian
This Anthropic engineer ( #RLanceMartin ) recently shared his full guide on how to run /loops with Fable 5.
If you're new to loop engineering and want to run fully autonomous Fable workflows, take notes
If you're new to loop engineering and want to run fully autonomous Fable workflows, take notes
✍1
🇨🇭 Global One SI - Prompt
GMKtec EVO-X3 launches July 6. 3rd-Gen Local AI Supercomputing Mini PC. AMD Ryzen AI Max+ 395 APU. 4TB stock storage. Native offline Qwen3 235B support. OCuLink eGPU expansion. GMKtec Claw AI Suite. The new EVO-X3 is getting ready for release. What eGPU…
If anyone is looking to build a multi-card rig for Local AI, current price ranges for working and new boards on Ebay show price stabilization. Deals can be found.
🔹 RTX-5090 (32GB) $3,599 - $5,199
🔹 RTX-4090 (24GB) $2,099 - $4,499
🔹 RTX-3090 (24GB) $1,100 - $2,000
🔹 RTX-5090 (32GB) $3,599 - $5,199
🔹 RTX-4090 (24GB) $2,099 - $4,499
🔹 RTX-3090 (24GB) $1,100 - $2,000
👀1
🇨🇭 Global One SI - Prompt
How to build a self-evolving AI brain with Hermes agent x Obsidian
This is the prompt Yale and Univ of Chicago researchers used when asking LLMs for new research ideas.
Feed LLMs prior work, ask for ideas, then measure how repetitive the ideas get.
The surprising finding is that LLMs often treat research ideation as connecting what already exists, while humans use a wider set of problem-finding moves.
LLM-generated ideas reveal a bias toward safe bridge-and-combine proposals.
Feed LLMs prior work, ask for ideas, then measure how repetitive the ideas get.
The surprising finding is that LLMs often treat research ideation as connecting what already exists, while humans use a wider set of problem-finding moves.
LLM-generated ideas reveal a bias toward safe bridge-and-combine proposals.
✍1
Compute is the new gold
Goldman sachs estimates $7.6t in ai infrastructure spending between 2026 and 2031.
Rhe breakdown:
1. compute...$5.1t...~67%
2. data centers...$2.15t...~28%
3. power...$358b...~5%
Annual ai capex could grow from $765b in 2026...
to $1.64t by 2031.
What does this mean?
Ai is becoming the largest infrastructure war in tech history.
Goldman sachs estimates $7.6t in ai infrastructure spending between 2026 and 2031.
Rhe breakdown:
1. compute...$5.1t...~67%
2. data centers...$2.15t...~28%
3. power...$358b...~5%
Annual ai capex could grow from $765b in 2026...
to $1.64t by 2031.
What does this mean?
Ai is becoming the largest infrastructure war in tech history.
Central banks just doubled their gold buying in a single month.
The World Gold Council reported net central bank gold purchases of 41 tonnes in May, more than double April's 19 tonnes.
Poland was the biggest buyer, adding 18 tonnes in May, bringing its 2026 total to 64 tonnes. Its reserves now stand at 614 tonnes.
China added 10 tonnes, its biggest monthly purchase since December 2024 and its 20th straight month of buying.
Total Chinese reserves now sit at 2,331 tonnes.
Turkey was the only major seller, offloading 3 tonnes in May, extending its 2026 net sales to 81 tonnes as it draws on reserves to defend the lira.
The 2026 Central Bank Gold Reserves Survey shows 89% of central banks expect global gold reserves to increase over the next 12 months.
A record 45% plan to increase their own holdings.
Gold gained +2% last week after four straight weekly declines but remains -25.4% below its peak.
The World Gold Council reported net central bank gold purchases of 41 tonnes in May, more than double April's 19 tonnes.
Poland was the biggest buyer, adding 18 tonnes in May, bringing its 2026 total to 64 tonnes. Its reserves now stand at 614 tonnes.
China added 10 tonnes, its biggest monthly purchase since December 2024 and its 20th straight month of buying.
Total Chinese reserves now sit at 2,331 tonnes.
Turkey was the only major seller, offloading 3 tonnes in May, extending its 2026 net sales to 81 tonnes as it draws on reserves to defend the lira.
The 2026 Central Bank Gold Reserves Survey shows 89% of central banks expect global gold reserves to increase over the next 12 months.
A record 45% plan to increase their own holdings.
Gold gained +2% last week after four straight weekly declines but remains -25.4% below its peak.