Most pharma or biotech have access to cryo-EM one way or another. But there is still a lot of room for new drug discovery paradigms based on cryo-EM and VCs know that. Could your research be the next Septerna, Gandeeva or MOMA? I write about it here https://t.co/pBKe97BfHf
DeepSeek will create a moment where all the AI startups that bought thousands of NVIDIA chips will probably go bankrupt, leading to a huge flow of NVIDIA GPUs into the second-hand market. This is in addition to startups whose model was to operate datacenters (e.g. Coreweave) and rent NVIDIA GPUs to other companies in hopes of generating ROI.
Lastly, the Magnificent 7 will finally start to slow down or cut any future orders from NVIDIA as there are already thousands, if not more, GPUs stored in warehouses and not fully utilized.
Then, everything about NVIDIA will start to unravel.
Presenting VirtualIce: Generate half-synthetic cryoEM micrographs; helps researchers visualize new proteins & diagnose issues (eg. preferred orientation, aggregation) and provides developers with ground-truths for ML dev (eg. picking, heterogeneity).
https://t.co/F0gbuohgfA
100,000's of our #cryoEM Autogrid Rings have been used since 2023 without issue. They're inspected w/multiple processes, including a visual inspection of every ring by experts.Our passion to make the best rings is evident in quality&package!See them #MM2024@MiTeGen@TedPellaInc
@sonjawelsch@IsraelF96135088 All top 50 pharma have some form of access to cryo-EM. About half of them have in-house access. Top 100-200 biotechs use service providers.
@RubinsteinJohn Cryo-EM Twitter feels a bit dead as of late, particularly the last 3-4 weeks. Maybe more (be it fake) likes will revive it / keep it alive?
This is actually democratising cryoEM! Brilliant work by @HariVenugopal2@CryoEM_Monash @GeorgRamm
Sub 3 angstrom on a 120 kV technai with a LaB6 source!
Read it! It's a masterpiece
https://t.co/J8ZJeA0DMS
AMAZON FINALLY GOES NUCLEAR
Amazon just bought a massive data center under construction right at one of America's best nuclear plants.
Operating costs for the Susquehanna nuclear plant in Pennsylvania are crazy low - likely around $25 per MWh.
Why is this a huge nuclear signal?
Until recently, Jeff Bezos was giving immense funding to enviro org NRDC, a group that has been trying to destroy American nuclear power plants.
NRDC succeeded at Indian Point, so New York City now runs on just natural gas and diesel. They almost succeeded at Diablo Canyon in California.
Amazon is bigger than Jeff, true. And Bezos may not have known the America-sabotaging dirty energy mission his money was supporting at NRDC, which only cut their anti-nuclear program late last year according to insiders.
But unlike Microsoft which has long had a pro-nuclear founder in Bill Gates, a stated interest in purchasing clean nuclear electricity, and even a nuclear power program, Amazon has been silent on nuclear.
Even Google has been louder than Amazon, saying 24/7 clean energy from nuclear would probably be an important clean energy resource in the future.
So by buying a data center that uses nearly HALF of a massive nuclear plant's output, Amazon is effectively announcing that they're a nuclear-powered company.
This is colossal for the "Nuclear Energy as ESG" story.
The major tech companies, which are essentially all entirety of the ESG sector, were avoiding buying nuclear power, instead misleading the public for years about using 100% renewable even though they knew it wasn't true.
Now, in order to accept the tech giants as ESG, major financial groups and the ESG ratings agencies that work with them will have to accept that these companies are openly powering themselves with zero-emission nuclear energy.
With the power demand for data centers soaring, and the major companies like Microsoft and Amazon now openly buying nuclear power to meet their needs, we are in the beginning of a new nuclear age.
Will the nuclear industry be able to meet this demand? That's the next challenge.
@SjorsScheres@nvidia I have it from good sources that cryoEM amounted for a sizeable amount of GPU usage in pharma companies for the longest time. Changing a little bit as of late because of AI.
Useful to distinguish at least four general choices:
1. Local IT
2. DYI shadow IT (necessary evil given #1)
3. Cloud equivalent of #1
4. Fully managed cloud services by specialized providers (e.g. Clovertex)
@HiCryoEM Beautiful! It is great to see ppl still building and proud of their own workstations instead of going cloud (or to local IT services which never understand the codes/jobs the WS will run). Reminds of my early #compchem days! :-)