Pranav

491 posts

Pranav

Pranav

@IamPranavJ

Engineer | Energy + AI | Previously @nasa

Katılım Ağustos 2012
958 Takip Edilen655 Takipçiler
Pranav
Pranav@IamPranavJ·
@kevinxu humanoid robots running 48 hours straight is the demo. running 48,000 hours with 99.9% uptime is the product. industrial robotics has been at the 99.9% bar for 40 years. that's the bar humanoids need to clear.
English
1
0
0
60
Pranav
Pranav@IamPranavJ·
@bindureddy 30x cheaper is the underrated number. inference cost is what drives deployment. benchmarks matter less when one model is 30x cheaper at similar capability.
English
0
0
0
46
Bindu Reddy
Bindu Reddy@bindureddy·
China is building pragmatic open-source AI models and no one can really stop them.... These models can already handle 50% of every day tasks effectively and are 30x cheaper to run In a few months, they will be able to handle most professional tasks - you don't need a 10T param model to automate work The US should stop complaining and catch up
English
28
7
77
4.6K
Pranav
Pranav@IamPranavJ·
@orbital_station 50 gas turbines because grid interconnection is a 5+ year wait. AI training can't sit in queue. onsite gas is the only way to compress that timeline.
English
0
0
1
5
Pranav
Pranav@IamPranavJ·
@TheAhmadOsman local AI on AMD is way underrated. ROCm has caught up enough that you can run 70B+ models on consumer cards. nobody talks about it because NVDA owns the marketing budget. but the hardware works.
English
0
0
0
114
Pranav
Pranav@IamPranavJ·
@aparanjape battery costs are dropping fast. but 24/7 solar+battery still breaks in winter and high latitudes. nuclear and firm power don't go away. they just compete differently.
English
0
0
0
7
Pranav
Pranav@IamPranavJ·
@sundeep this chart says it all. data centers use about 0.02% of US water. one alfalfa field consumes more than a hyperscale facility. and yet.
English
0
0
0
11
sunny madra
sunny madra@sundeep·
I love almonds (and data centers)
sunny madra tweet media
English
4
1
27
1.8K
Pranav
Pranav@IamPranavJ·
@KobeissiLetter 8.6M sounds huge. US burns 20M barrels a day. one week of SPR drawdown is about 10 hours of consumption. SPR was built for supply shocks. now it's a price tool.
English
0
0
0
43
The Kobeissi Letter
The Kobeissi Letter@KobeissiLetter·
US oil reserves are falling at a record pace: The Strategic Petroleum Reserve (SPR) dropped -8.6 million barrels last week, the largest weekly drawdown on record, per Zerohedge. This marks the 7th consecutive weekly decline, the longest streak since 2023. Over this period, US oil reserves in the SPR have dropped -31 million barrels, to 384 million, the lowest since October 2024. This wipes out nearly half of all the oil added to the SPR between July 2023 and March 2026. Meanwhile, US crude oil exports jumped +742,000 barrels last week, to 5.5 million barrels per day, the 2nd-highest in at least 3 years, after a record 6.5 million barrels per day 2 weeks prior. The world is scrambling for American oil.
The Kobeissi Letter tweet media
English
156
456
1.7K
158.6K
Pranav
Pranav@IamPranavJ·
@tbpn @esaagar the populism is real. most of the complaints are misinformed. data centers don't pollute air. they strain grids. fix the grid and the politics resolves itself.
English
0
0
0
60
TBPN
TBPN@tbpn·
Breaking Points' @esaagar says AI companies are "fighting against a very, very big force" in the US. "There's this rising populist tide against the data center movement. And against Abundance-style assurances from politicians and companies." "Something is happening. The tide has not only turned, but coming." "People just feel like this is out of their control. About AI especially. They're like, 'I want impact. I want a say.'" "There's an overwhelming animus, and people are angry. It's about a lack of control. 'You're coming to take my job, you're increasing my electricity prices, you're changing my nation. And I have to have a say as a citizen.'" "It's no longer just about electricity, it's about the whole picture." From his appearance on the show last month.
Garry Tan@garrytan

Sanders and AOC introduced a bill to pause ALL AI data center construction. 300+ local bills filed. Half of planned 2026 data centers facing delays or cancellation. Each one brings billions to local economies. The people who say they want American jobs are trying to block the biggest job creation engine since the interstate highway system.

English
11
5
58
18.9K
Pranav
Pranav@IamPranavJ·
@EMostaque yeah TAM is way bigger. but humanoids stack every AI bottleneck on top of physical manufacturing. compute scales fast. atoms don't. you're looking at a 2030+ economy effect.
English
0
0
0
157
Emad
Emad@EMostaque·
Spent some time this week looking at the economic impact of humanoid robots and its so much bigger than digital AI The numbers get very crazy, has anyone dug into it and not found this the case?
English
11
1
32
4.4K
Pranav
Pranav@IamPranavJ·
India is adding 570 TWh of new electricity demand by 2030. that's the entire grid of germany. coal will still be 68% of supply in 2026. western investors keep talking about solar. coal is doing 7 of every 10 watts.
Pranav tweet media
English
0
0
0
16
Pranav
Pranav@IamPranavJ·
@NoLimitGains 5.11% on the 30Y changes the math for long-cycle infrastructure. nuclear plants amortize over 50 years. data centers over 15. the financing gap just widened on every AI buildout plan.
English
3
0
3
1.2K
NoLimit
NoLimit@NoLimitGains·
Oh shit
NoLimit tweet media
English
161
75
1.1K
167.9K
Pranav
Pranav@IamPranavJ·
@ai_for_success matches what i see. claude has been pulling ahead for engineering work for 6 months. openai owns consumer. anthropic owns builders. those are different races.
English
0
0
0
31
Pranav
Pranav@IamPranavJ·
@is_OwenLewis agree on the framing. but the pro-data-center side keeps overclaiming. data centers DO evaporate water. AWS oregon: 5M+ gallons a day. just own the numbers.
English
0
0
0
11
Owen Lewis
Owen Lewis@is_OwenLewis·
Not building datacenter is economic self-sabotage. It's also civilizational sabotage, but let's stick to the level most people can relate to. If there's a datacenter built in your area, that means a massive increase in tax revenue flowing into local coffers, which can be used to improve the lives of local residents. It'll mean a few extra jobs, and (because datacenters now have to bring their own power) more abundant available energy, likely at cheaper prices. It's relatively quiet after construction is done, and the only byproduct is heat. They use some water, but not much. Concerns over water have been greatly overblown due to fearmongering and outright lies. There are zero significant downsides. How often can that be said for anything this big? Opposition to datacenters is the height of irrationality, and I just hope enough people realize the benefits before they're all forced up into space.
Payton Alexander@AlexanderPayton

Data centers are generating trillions of dollars in value for the economy. Data centers don’t consume water, they return it to the environment. Data centers lower household energy costs by producing their own electricity and generating stable demand that reduces operating costs for the grid. Data centers create jobs for construction workers, electricians, and other skilled workers year round, all across the country. Data centers data centers pay billions in state and local taxes without using public services, meaning more resources for the rest of us. Data centers power every part of the digital economy in ways most people don’t even know, even enabling things as mundane as stocking products at grocery stores. Data centers are driving the AI revolution that will keep America ahead of China and enable universal high income for everyone.

English
117
53
422
28.1K
Pranav
Pranav@IamPranavJ·
@lakshmishaks india's GPU import bill hits $10B+ this year. assembly is the easy part. packaging and advanced fab are 10-year gaps. and we're not on that timeline.
English
0
0
0
2
Lakshmisha K S
Lakshmisha K S@lakshmishaks·
India’s IT Hardware PLI may soon expand beyond laptops/PCs into AI infrastructure. MeitY officials say AI servers could be added in the next revamp as India pushes for local AI hardware manufacturing amid soaring global GPU/server demand. From assembly → compute stack.
Lakshmisha K S tweet media
English
5
75
267
4.5K
Pranav
Pranav@IamPranavJ·
@JigarShahDC agree on nuclear. timing is the harder problem. AI compute is on a 3x curve in 5 years. new reactors take 7-12. the gap gets filled by whatever firm power is on the grid today.
English
0
0
0
8
Jigar Shah
Jigar Shah@JigarShahDC·
Everyone's asking: what actually powers AI data centers? The answer is more complicated — and more interesting — than "natural gas vs renewables." Nuclear is always the best... A 🧵 on what's really happening, and where it's going.
Duncan S. Campbell@duncancampbell

@isabelleboemeke Look at the right of the chart. This can be 90%+ clean for same price as TMI restart. And TMI won’t offer 100% clean baseload either. More like 93% with gas backup! But to be very transparent we also show the full range of tradeoffs.

English
7
8
44
15.4K
Pranav
Pranav@IamPranavJ·
meta: 6.6 GW of nuclear. microsoft: restarting three mile island. amazon and google: PPAs signed. five years ago you couldn't get a utility to talk to you about new reactors. now there's a queue.
Pranav tweet media
English
0
0
0
10
Pranav
Pranav@IamPranavJ·
@rauchg agents amplify whatever skill you bring. weak skills + agents = more bad work shipped faster. strong skills + agents = absurd leverage. people skip the first half and wonder why.
English
0
0
0
391
Guillermo Rauch
Guillermo Rauch@rauchg·
If you become exceptional at managing agents, but are also exceptional in your understanding of the fundamentals, you will be unstoppable. We all prefer to work with masters of their craft. What’s new: you can’t afford to miss out on the amplification agents have on your output
English
90
79
1.4K
54.7K
Pranav
Pranav@IamPranavJ·
@ARK_Funds you left off the biggest one. power. grid build can't match what AI compute will need by 2028. the other three are solvable. electrons aren't.
English
0
0
0
15
ARK Funds
ARK Funds@ARK_Funds·
Analysts estimated that global IT spending would hit $5.4 trillion in 2025, with AI infrastructure demand driving much of that growth, and even more this year. Where do you think the biggest bottleneck is?
English
3
6
11
9K
Pranav
Pranav@IamPranavJ·
@PolymarketMoney 1000x compute = 1000x power. the grid took 100 years to build. we don't have 100 years. and nobody is on a 10-year build pace either.
English
0
0
0
30
Polymarket Money
Polymarket Money@PolymarketMoney·
$NVDA CEO Jensen Huang says computing may need “1,000x more” energy than we currently have.
English
8
9
99
7.2K
Pranav
Pranav@IamPranavJ·
@riteshmjn yeah and it gets worse. one engineer plus 24/7 production inference costs the same as a senior at a top firm. but ships 10x the throughput.
English
0
0
0
120
Ritesh Jain
Ritesh Jain@riteshmjn·
Unpopular opinion but slowly people have started talking about it. Cost of compute ( not subsidized) is significantly higher than the starting salary of an Indian IT worker.
Ritesh Jain tweet media
Ritesh Jain@riteshmjn

@iamshalabh21 Completely agree that’s why I have turned bullish on Indian and global software.

English
23
43
308
16.7K