Jonas Hernlund

329 posts

Jonas Hernlund banner
Jonas Hernlund

Jonas Hernlund

@jonashernlund

AI is infrastructure, not software. Writing about power, chips, and the physical systems that decide the outcome. SVP Autonomy and Energy at Scania. Ex-Google

Stockholm, Sverige Katılım Nisan 2009
326 Takip Edilen122 Takipçiler
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@elonmusk Mass drivers solve delta-v. The binding constraint is lunar ISRU. You need fission on the surface to power the launcher and refine regolith into solar collectors. Path to petawatts runs through reactors before kinetic launchers.
English
4
3
21
2.3K
Elon Musk
Elon Musk@elonmusk·
Path to Petawatts is Mass drivers on Moon
English
5.2K
3.6K
48K
22.1M
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@sama Underrated because most teams haven't paid the production tax yet. Tool retrieval, idempotency, partial-failure recovery, and run-level observability are where the SDK either compounds or breaks. Demo agents are easy. Long-running ones expose the plumbing fast.
English
0
0
2
2.1K
Sam Altman
Sam Altman@sama·
Agents SDK 2.0 is underrated
English
388
74
2.3K
211K
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@Kantrowitz Ignored because the eval layer never caught up. Image gen is still measured by cherry picked demos and arena votes. Real progress hides until somebody ships it into a paid workflow. The leap isn't small, the measurement is.
English
0
0
1
52
Alex Kantrowitz
Alex Kantrowitz@Kantrowitz·
Hard to put into words just how much better generative AI images have gotten. Massive leap just happened and it's mostly being ignored.
English
3
1
19
1.7K
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@KobeissiLetter Capex headline hides the real bottleneck. Constraint isn't dollars, it's grid interconnect queues and transformer lead times of 3 to 5 years. Capital is liquid, electrons aren't. Whoever locked in power contracts in 2024 captures most of this 2026 spend.
English
0
0
0
161
The Kobeissi Letter
The Kobeissi Letter@KobeissiLetter·
Big Tech CapEx has reached unprecedented levels: The combined CapEx of Amazon, $AMZN, Google, $GOOG, Meta, $META, and Microsoft, $MSFT, is expected to surge +98% YoY, to a record $715 billion in 2026. This is nearly 3 TIMES the amount spent in 2024 and more than 5 TIMES 2023 levels. Amazon now expects CapEx to approach a record $200 billion this year. This is followed by Google and Microsoft which are both guiding $190 billion in CapEx for 2026. Meta has also raised its expectations by +$10 billion to $125-$145 billion. To put this differently, each company is projected to spend nearly as much in 2026 as they did in the previous 2 years combined, or more. The AI Revolution is set to accelerate.
The Kobeissi Letter tweet media
English
183
253
1.6K
223.7K
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@paulg Truer than that. Legacy orgs cant describe what they want because their workflows never produced structured artifacts. New companies bake it in day one. The data shape moat compounds before the model layer matters. Gap is structural not skill.
English
0
0
5
671
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@gdb Funny version of a real shift. Agentic coding moves the bottleneck from keys per second to PRs reviewed per hour. Single monitor works because most of the screen is now diff and test output. The IDE that nails parallel agent review at scale wins the next round.
English
1
0
1
54
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@Andercot Footprint is even worse than that. Desert solar capacity factor is ~25%, so true 24/7 backing needs 4x panels plus battery overbuild. SMRs do 300MW on five acres. For GW-scale AI loads the energy density math only works with fission. Solar is for daytime arb.
English
1
0
0
2.8K
Andrew Côté
Andrew Côté@Andercot·
This is the footprint ratio of data center to solar panels in the sunniest country in the world. Yeah, I think we're gonna have to go nuclear.
Andrew Côté tweet media
Object Zero@Object_Zero_

This 100MW data center in UAE is the largest solar powered datacenter in the world. There are currently 1,300 data centers in the world that are bigger than this one, but this one is the largest solar powered one. That’s 10 square kilometres of solar panels you can see. The datacenter itself is 0.02 square kilometres, so a solar powered datacenter is ~500x larger than a data center using any other form of power. A five hundred times larger site. UAE has some of the highest solar irradiance anywhere on Earth, it is an inhospitable desert. Averaging 9.7 hours of sunlight per day with average irradiance above 2,200 kWh/m^2. If you build this somewhere else, you need more solar panels because your irradiance will almost certainly be lower. Even if the world had an infinite supply of free solar panels, solar power will not be free. Anyone who has ever done major capital projects, who looks at where data centers need to be in the next 5 years and the next 10 years… we know it aint solar. Sorry. You struggle to even build a train track that’s 100 miles long and 10ft wide anywhere in the West, there is zero chance of build 100 square mile solar farms for GW compute. This is why people are talking about space compute. Deploying into space is one strategy to solve the constraints. But there are faster and more scalable strategies, that get you to mass deployment of multi GW data centers. There are strategies that also allow you to power the 10 billion robots and their newtonian actuators, that immediately follow the inference demand cycle. Step back and look at the full cycle of this industrial revolution… There will be billions of chips, but there will be trillions of actuators. This biggest part of this revolution is the embodiment cycle, and it’s big by a factor of 20 or 50x over the stuff that comes before it. There is no analogy in human history for the scale of this economy, of the demand it will place on energy and commodities. The humans own the Earth, and if you exist inside their legal system, they won’t let you turn the surface of their planet into glass. But they do want your chips and your actuators to serve their needs and desires. There is a way to do all of this, and so it will happen.

English
534
1.1K
9.2K
1.5M
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@levie Engineering parity is real. The bottleneck moves to what AI doesn't accelerate. FDA validation, GxP data lineage, change control on validated systems. Life sciences hires the same engineers but still ships at regulatory cadence. New rate-limiter is data ops and approval cycles.
English
0
0
0
249
Aaron Levie
Aaron Levie@levie·
If you think AI replaces software engineers, here’s a quick thought experiment. Imagine you’re a life sciences company. 10 years ago you want to invest heavily in lab automation, processing data at scale, and other software. You look at the cost of doing so and realize you can’t compete with tech for as many engineers as you need, so you pare down your goals and do what you can. Every new software project has a fixed cost of a certain sized team, so you can only do so much given budgets, ability to compete for talent, and other trade offs. Now, AI comes along. And all of a sudden you have the *exact same* output tokens as the best tech companies in the world. Your engineers are using the same AI models as the tech industry, which means you have just boosted your engineering team by a some meaningful amount, while also neutralizing your differences with tech. Do you continue with your pared down approach, or do you start to hire more engineers because each engineer is 2X or 5X more capable than before? In almost every company I’m talking to, they’re doing the latter. Now extrapolate this to every bank, manufacturer, industrial company, retailer, and on and on. And extrapolate it not to just large enterprises, but also every SMB up and down the stack of these value chains. Oh, and also extrapolate this to other job functions, not just engineers. Resource scarce domains in marketing, legal, finance, design, and so on. If you’re wondering why new jobs show up because of AI this is the reason. Any other view of what happens doesn’t contemplate the variety of unmet needs there are in the economy.
unusual_whales@unusual_whales

Nvidia CEO Jensen Huang: “The narratives of AI destroying jobs is not going to help America: it's false."

English
71
77
419
82.2K
Jonas Hernlund
Jonas Hernlund@jonashernlund·
Google Cloud margin tripled in 12 months: 9.4% → 32.9%. Revenue +63% YoY to $20B. Capacity-constrained. Backlog $460B. Big tech AI capex 2026: $725B (+77% YoY). The bubble crowd just lost the unit economics debate. Custom silicon is the moat.
English
0
0
4
85
Jonas Hernlund
Jonas Hernlund@jonashernlund·
Running autonomous trucks taught me the hard part is not the AI. It's who gets paged at 2am when one is stuck in mud and the customer is short 8 trailers by sunrise. Most AI startups still don't staff for that.
English
0
0
1
34
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@pmarca Market cap built on top is not the same as margin captured. Open weights commoditize the model layer. Inference and tooling still flow to neoclouds and hyperscalers. Tinkerers eat last. The capital stack has not moved as much as the leaderboard.
English
0
0
0
62
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@gdb Logged-out to logged-in via image gen is the real story here. Image becomes the cheapest acquisition surface for chat products. Whoever ships image first in a workflow owns the funnel. Text-only competitors lose entry traffic they never see.
English
0
0
0
350
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@sama Depends on the workload. Agents running long tool loops live or die on token cost and latency. One-shot reasoning rewards intelligence. The dichotomy hides the real product split. You are optimizing for two different things, not one.
English
0
0
0
101
Sam Altman
Sam Altman@sama·
i keep thinking i want the models to be cheaper/faster more than i want them to be smarter but it seems that just being smarter is still the most important thing
English
2.3K
351
12.1K
948.2K
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@firstadopter Utility isn't model quality. It's IDE penetration plus how far the agent gets in your repo before hitting permission errors. Cursor and Claude Code won the surface area war in 2025. Gemini needs a distribution story before it needs another benchmark win.
English
0
0
0
267
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@levie Specialization isn't just market driven, it's liability driven. SOX, FDA, OSHA need a human owner of record on every workflow. Agents can write the PRD and the code, but until a model can sign the deploy and own the audit trail, the boundary stays.
English
0
0
1
203
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@tobi This is the format that wins. Browser agents clicking through checkout flows is theater. When buyer agents talk directly to seller agents, attribution breaks, retargeting dies, and most performance marketing ages out fast.
English
0
0
2
827
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@pmarca Aggregate is up but the mix is shifting hard. Senior architects who review agent output are growing fast. Junior pure-coding roles are quietly compressing. One senior now ships what used to need three. Indeed numbers won't show this for four quarters.
English
1
0
0
279
Jonas Hernlund
Jonas Hernlund@jonashernlund·
@firstadopter Huang sells picks and shovels regardless of who wins or how many jobs vanish. Of course he's bullish. The actual displacement signal is lagged and quiet. It lives in OEE numbers two quarters after a line goes agentic, not in earnings calls.
English
0
1
3
1.4K
tae kim
tae kim@firstadopter·
Nvidia CEO Jensen Huang says don't listen to CEOs with a god complex on AI (lol, Dario) $NVDA On AI destroying jobs: "these kind of comments are not helpful .. somehow they became CEOs, you adopt a god complex and before you know it, you know everything" "ground ourselves to talking about the facts" AI will "generate hundreds of thousands of jobs .. trillions of dollars [to the U.S. economy]"
English
320
609
5.6K
1.2M
Department of War CTO
Today, the @DeptofWar entered into agreements with SEVEN of the world's leading frontier AI model and infrastructure companies to deploy frontier capabilities on the Department's classified networks: • SpaceX • OpenAI • Google • NVIDIA • Reflection • Microsoft • Amazon Web Services This is just the latest initiative in our mandate to create an AI-FIRST WAR DEPARTMENT 🇺🇸
Department of War CTO tweet media
English
391
1.1K
4.9K
705K