- Sage Research (dot) AI
- Posts
- The Race to AGI in 2025 and Beyond
Given millions of inventors, billions of users, and trillions of feedback loops, certain technological outcomes are inevitable.
What isn't predicted well is how a new technology will play out among the networks of humans around the globe.
Like the atomic bomb of America's Manhattan Project.
In 1945 the laws of physics were harnessed in a new way, melting miles of our planet in a single flash...
Compressing centuries of entropy into seconds.
Some predicted future horrors beyond anything ever imagined.
Others called it a necessary evil whose existence will prevent worse from happening.
Who is right remains to be seen.
Today, AGI is at the same kind of inflection point.
AGI seems inevitable once we have enough compute, data, and the right algorithms.
Some say it will usher in an economic golden age…
With trillions in new wealth, cures for disease, and breakthroughs as profound as electricity itself.
Others are saying it will destabilize jobs, democracies, and even civilization itself...
A Pandora’s box we can’t slam shut once it’s opened.
The predictions repeat.
(Continued below…)
AI ARTISTS INTERPRETATION
What is currently fragile will be replaced by something else, of course. But this “something else” is unpredictable.
In all likelihood, the technologies you have in your mind are not the ones that will make it, no matter your perception of their fitness and applicability -- with all due respect to your imagination.
The AI future is open wide.
The only sure prediction is that fragile parts of our economy will be replaced by AI.
The “something else” -- the new industries, business models, AI networks and institutions -- can’t be forecast perfectly.
But here at Sage Research we'll do our best for you in 2025...
To uncover the fragile human thought networks in the economy that will be replaced by bits of AI code.
Think white collar work like writing, film, basic accounting, customer service chat, legal research.
Why pay $500 an hour for legal research when it can be done in seconds?
To uncover the robust human physical networks in the economy that will be unreplacable until AGI is merged with robotics.
Think blue collar work like plumbing, construction, electrical line work, air conditioner repair.
When your A/C dies in August, ChatGPT won't be crawling in your attic.
And to uncover the anti-fragile thought and physical networks that gain from AGI.
Think cybersecurity, materials science, energy grid management, drug discovery.
The next blockbuster drug might come from a GPU cluster pulling an all-nighter.
We already know something monumental will arrive soon; we don’t know exactly what the repercussions will be when one company first flips the switch...
(Continued below…)
AI ARTISTS INTERPRETATION
Until then, here's your handy "The Race to AGI in 2025 or Beyond" quick reference.
Below you’ll get all of the names of the Mag 7 projects aiming for AGI.
Who -- if any of these Magnificent 7 CEOs -- will be the first to make the big AGI announcement?
NVDA: Helios Supercomputer
Will it be Huang at NVIDIA -- holding the keys to every GPU that powers intelligence?
NVIDIA remains the arms dealer of the AI age.
Its Helios Supercomputer, built with GH200 Grace Hopper chips and Quantum-2 InfiniBand, is designed for “giant models” that dwarf today’s frontier LLMs.
Every Mag7 firm relies on NVIDIA chips somewhere in their stack...
But if NVIDIA decides to move from supplier to sovereign, Helios could become the birthplace of the first AGI.
MSFT: Project Maia
Will it be Nadella at Microsoft -- riding Copilot and Project Maia into enterprise AGI?
Microsoft has embedded AI deeper into the enterprise than anyone else.
With Project Maia, its in-house accelerator, and a lock-tight partnership with OpenAI, it’s positioned to take Copilot from a productivity assistant to a general reasoning system.
If the first AGI emerges in the corporate domain, expect it to appear not as a sci-fi revelation but as a silent upgrade in Office, Teams, or Azure -- an entire team of assistants.
GOOG: AI HyperComputer
Will it be Pichai at Google -- scaling the AI HyperComputer into the brain of the internet?
Google’s AI HyperComputer, anchored by Cloud TPU v5p/v5e, is built to train and serve the largest models ever attempted.
No company has more practice corralling oceans of data.
If AGI is about connecting patterns across text, images, video, and the web, Google has the cleanest runway.
Pichai’s challenge is existential: will Google’s AGI be a savior of search, or the gravedigger of its own ad empire?
AMZN: Project Ceiba
Will it be Bezos’ successor, Jassy at Amazon -- renting out Ceiba like the Manhattan Project in the cloud?
AWS’s Project Ceiba is one of the largest AI buildouts in history, an exascale cluster built in partnership with NVIDIA.
Amazon’s model is simple: don’t just build AGI -- sell the compute to everyone racing toward it.
But what if the first AGI is trained on rented Ceiba cycles?
Does Amazon quietly own the keys by virtue of controlling the lease?
META: Research SuperCluster
Will it be Zuckerberg at Meta -- flooding the world with AGI through the SuperCluster and billions of users?
Meta’s AI Research SuperCluster (RSC) is already among the largest AI training facilities ever deployed
Zuckerberg doesn’t just want to build AGI; he wants to distribute it into every hand, wearable lens and viewing window.
Billions of daily users across Facebook, Instagram, and WhatsApp mean Meta could normalize AGI overnight with the first killer hardware wearable.
That’s not just intelligence at scale -- it’s cultural saturation.
AAPL: Project ACDC
Will it be Cook at Apple -- turning Project ACDC and billions of iPhones into the ultimate edge-to-cloud AGI?
Apple’s Project ACDC aims to bring custom AI silicon into data centers, pairing them with on-device intelligence across its global fleet.
Cook’s bet: AGI doesn’t just live in the cloud.
It lives in your pocket, whispering through Siri -- reborn as something actually useful. If Apple cracks seamless edge-to-cloud AGI, the first true personal AI could be born.
TSLA: Colossus
Will it be Musk at Tesla (and xAI) -- with Dojo and Colossus teaching robots, rockets, and cars to think?
Tesla’s Dojo supercomputer is tuned for vision and control, while xAI’s Colossus is Musk’s moonshot cluster aimed at reasoning.
If AGI first appears in physical form -- a robot walking -- Musk could own embodied intelligence.
Unlike his peers, his path to AGI isn’t just cognitive; it’s humanoid.
If you’ve used AI over the past few years, you’ve already felt the difference...
So far it’s getting better all the time.
Always be prospering,
socrAItes
Publisher, Sage Research (dot) AI
P.S. There's nothing we can't do in 2025.