Natural Cooling, how it works and its disappearance
Turbine ventilators were prevalent in the past
As we enter the summer in North America, things to think about
@Barchart Wrong , what your stating is the first derivative ( underperformance by largest margin ) and what youโre showing in the chart is relative price .. you need to take the first derivative with respect to time for your statement
NVIDIA MADE A $249 DESK CHIP THAT TURNS LOCAL LLMS INTO A TINY PRIVATE BACKEND
jetson orin nano super is basically a pocket-size ai board: 67 TOPS, 1024-core ampere gpu, 8gb memory, 7-25w power draw. not a gaming pc, not a rented cloud box, just a small nvidia dev kit running llama, mistral, qwen and deepseek from your own table
that changes the cost structure. the same automations people casually run through openai for $100-200/month can move the boring work local: summaries, drafts, code checks, document q&a, scraping cleanup, private agents. the board is $249 once, then around $2/month in electricity
the setup is almost stupid. install ollama, pull a 7b model, aim your app at localhost. your workflow still talks to an openai-style api, but now the request never leaves the room and the bill does not grow every time a script loops
no, it is not replacing claude for the hardest 20%. but it absolutely attacks the expensive background layer people forgot they were paying for every day
cloud ai made people rent intelligence by the token. nvidia is quietly showing what happens when that meter sits under your desk instead.
Why are private credit lenders making loans to data centers? I asked one, here's what he said:
- Data centers w/ Mag7 as multi-year tenants have years of high-confidence revenue locked in
- collateral: chips/land
Apple https://t.co/sQ8X2r75Qu
Spotify https://t.co/8VFyhE07RZ
Ukraine's security service, @ServiceSsu, has released more info about Operation Spiderweb and a 4-min video showing new footage, including drones landing atop A-50 early warning and control aircraft.
This is the most important chart you'll see today.
It shows the U.S. Equity Risk Premium (ERP), more specifically:
S&P 500 earnings yield minus the 10-year Treasury yield.
In simple terms, it answers one critical question: ๐
Trump sued his own government for $10 billion. Today he settled with the DOJ he runs and cut the judge out of the process.
The deal: a $1.8 billion taxpayer-funded slush fund his Attorney General controls, with no court oversight and no transparency. Itโs the most corrupt scheme of his presidency.
The final point Rory cites from Iran in the below tweet is the most critical one.
Iran has been crystal clear since the conflict began that they are unwilling to give up possession of their ~440kg of 60%-enriched HEU. That's the central "risk" that Trump and Netanyahu started this whole war over. When Trump said Iran was "just weeks away from a nuclear weapon", what he really meant is that it would only take a few weeks to enrich that 60% HEU to 90% (weapons grade). That's what the whole war was started over in the first place, and it will always be the central issue that has to be resolved before the conflict can end.
It's plausible that Trump is giving up to save face, declare victory, and move on. In that case this whole conflict will have been for nothing. Far more likely, Trump is buying more "negotiations occurring under ceasefire" time to avoid congressional approval requirements. It's also plausible that Trump incorrectly believes that Iran will allow its 60%-enriched HEU to leave Iran, and when he finds out they refused (consistent with exactly what they've repeatedly said they would do), he may accuse them of breaking the agreement.
It's crystal clear that Iran sees that outcome coming, and is "in damage control mode" now, making it publicly clear that they have not agreed to hand over their HEU. They don't expressly say here that they will never turn it over, but they've made clear in other statements in the last few days that this exact order was given by the Supreme Leader this past week.
I therefore conclude 1 of 2 things has to be true: 1) This is another "fake news" peace deal that's intended to buy more time before facing congressional approval challenges, but will ultimately prove just as meaningless as the last several similar "peace deals" that never panned out, OR 2) Trump is giving up and will try and save face by declaring victory, taking credit for reopening the strait, and might then pivot to Epstein-files tactics to brush minor details like Iran still has the nuclear material that could be enriched to weapons-grade in just a few weeks and Iran will begin charging massive tolls for traffic transiting the Strait under the proverbial rug.
But #2 is very unlikely, because Rubio has already made public statements (Saturday, from India) reiterating the nuclear materials hand-over red line. He wouldn't be pushing that message if Trump was giving up and looking to save face.
I lean toward #1 above, considering that Iran has already emphasized that Trump's formative MOU "does not in any way mean free passage thru the Strait".
Remember, Trump needs to avoid any "resumption of the war" because he's already well beyond the 60-day deadline for congressional approval to continue the war. So this whole thing looks to me like a tactic to defer the nuclear materials negotiation (i.e. the real "red line" substantive matters that were the main reason the war started) into a 30-day drag-this-out phase, where we get at least some traffic through the strait and stay immune from congressional approval demands.
If that's right, it will prolong this until both sides eventually reach the obvious conclusion that neither is willing to bend on its red-line demands, which all center on the nuclear material hand-over that hasn't even been discussed in the negotiations that precipitated this weekend's "peace deal" announcement from @POTUS.
If you're tempted to say "But...They are talking about opening the Strait NOW, and that will solve the global energy supply problem, so this is still massively bearish oil prices", think again. Better yet, follow @CommodMkt. Even if the Strait were fully re-opened to pre-war traffic levels (very unlikely), it would take 6-8 weeks for the tankers to reach their destinations before energy supply could begin to recover. By then we'll have reached the next red-line challenge in this negotiation. Bottom line, this ain't over till it's over. And it's not over nor can it or will it end this weekend.
I'm working on a Substack post going into more detail on all this now. It will explain what the MSM have consistently failed to explain: Why the 440kg of 60% HEU is so important, and why there seems to be such a disconnect between Trump, Lindsay Graham and others insisting "Iran was only a few weeks away from a nuclear weapon" vs. Tulsi Gabbard insisting "Iran doesn't have and hasn't had a nuclear weapons program for over 20 years". Believe it or not, they're both right, and the explanation is nuanced. Follow my substack now if you want to be notified. Should be posted by early Sunday morning.
To finish on this, I think all hell breaks loose at or before we reach 6.5 bln barrel global visible inventories.
2.8 crude linefill and tank bottoms
1.3 crude and products on water (in transit)
2.4 products linefill and distribution channel
I'm a touch more optimistic that the JPMorgan / BBG publication, but not by much, at current deficits, shortages can't be avoided in three to four months.
Even sooner for products with low inventories and high operational floor caused by the long distances they travel from production to consumption regions, think petchems and certain light ends.
MICROSOFT CANCELED CLAUDE CODE!
IT COST TOO MUCH.
Major tech companies are confronting the steep reality of AI inference costs as the era of heavy subsidies appears to be ending.
Microsoft is canceling most internal licenses for Anthropicโs popular Claude Code tool by June 30, 2026 less than six months after rolling it out broadly to engineers primarily due to escalating token-based expenses, while shifting teams toward its own GitHub Copilot CLI.
This mirrors broader pressures: Uberโs CTO revealed the company had already exhausted its entire 2026 AI budget in just four months thanks to heavy Claude usage among thousands of engineers (with individual monthly costs often hitting $500โ$2,000), and GitHub is transitioning Copilot to usage-based billing with higher per-token rates starting June 1st.
The reality is good enough AI will expand and constantly get better removing the oxygen of the most expensive.