Home
Đăng nhập
Đăng ký
Nội dung học
Loading...
The Amazon Deal That Just KILLED Microsoft’s Future. - Video học tiếng Anh
Luyện nghe
Luyện nghe
/
Video
/
The Infographics Show
/
The Amazon Deal That Just KILLED Microsoft’s Future.
The Amazon Deal That Just KILLED Microsoft’s Future.
Chọn chế độ học:
Xem phụ đề
Chọn từ
Viết lại từ
Highlight:
3000 Oxford Words
4000 IELTS Words
5000 Oxford Words
3000 Common Words
1000 TOEIC Words
5000 TOEFL Words
Phụ đề (271)
0:00
Everyone thinks Microsoft won the AI war… but they may have just lost it overnight.
0:05
Because Sam Altman just made a $50 billion move with Amazon to stab Microsoft in the back… and
0:11
almost no one noticed. On the surface, it looks like another massive AI deal. But
0:16
hidden inside it is a shift that sidelines Microsoft and removes the one clause that
0:21
actually kept control in check. And that changes everything.
0:26
Because this isn’t just about building powerful AI anymore. It’s about who owns it, who controls it
0:32
and who’s willing to burn billions to get there. By the time most people realize what just
0:36
happened, the balance of power may already be gone.
0:39
Chapter 1 - The $50 Billion Betrayal To the public, OpenAI looks like one
0:44
of the biggest success stories of the modern era. A Silicon Valley startup
0:47
that turned into a household name almost overnight. It built tools used by millions,
0:52
pushed AI further than anyone expected, and wrapped it all in a mission to benefit humanity.
0:58
That was the illusion. In reality, OpenAI is a mess that’s bleeding cash.
1:02
In 2024, its projected revenue was approximately $3.7 billion.
1:07
Its losses? $5 billion. That’s like buying a mid-sized
1:11
airline and setting the whole thing on fire every 12 months. Just to keep the servers running.
1:16
That’s not a sustainable or successful business model.
1:19
That’s not the sign of a healthy and stable company. And it wasn’t a one-time thing.
1:23
It was a trend. In the first half of 2025,
1:26
figures revealed that OpenAI was losing extraordinary amounts of money, generating
1:30
around $4.3 billion in revenue while recording losses of up to $13 billion. Some estimates
1:37
suggest that the company’s total losses could exceed $140 billion between 2024 and 2029 alone.
1:44
But it’s not that surprising. Training frontier AI models isn’t cheap.
1:48
Neither are the salaries of leading researchers and computer scientists, or the construction
1:53
and operating costs of data centers. OpenAI is burning through money at breakneck
1:57
speed. Its entire business model is founded on the idea of convincing investors that someday,
2:03
somehow, all of this loss will be worth it. Microsoft bought into that idea.
2:07
It poured billions into OpenAI and secured what seemed to be an exclusive hold over the most
2:13
valuable AI startup on the planet. Microsoft’s CEO, Satya Nadella, even said: “It wouldn't
2:18
matter if OpenAI disappeared tomorrow. We have the data, IP rights, and all the capability.”
2:24
Nadella thought that, for all intents and purposes, he owned OpenAI.
2:28
He was wrong. In February 2026,
2:30
Sam Altman orchestrated an enormous $50 billion infrastructure deal with Amazon.
2:36
In doing so, he effectively ended Microsoft’s exclusive cloud rights.
2:40
That wasn’t supposed to happen. Microsoft was supposed to be the only serious
2:44
player in this game. Microsoft Azure was meant to be the default home for OpenAI’s technology.
2:49
Amazon was the rival. The company you compete against, not partner with.
2:54
This wasn’t just a new vendor agreement or some sort of multi-partner strategy;
2:58
it was OpenAI blatantly betraying the tech giant that helped build it.
3:03
The question is why? Why would Altman risk the
3:06
wrath of Microsoft? Why jeopardize what seemed to be the most powerful relationship in the industry?
3:11
Because the Amazon deal wasn’t just about getting more servers and resources.
3:15
It was a weapon, built for one mission… to defeat a more powerful enemy:
3:20
the United States government. Chapter 2 - The FTC’s Trap
3:25
For years, it looked like compute was going to be the biggest challenge OpenAI would ever face.
3:30
But the funding from Microsoft introduced OpenAI to something else: antitrust.
3:34
To observers and analysts - including those in government authorities,
3:37
such as the Federal Trade Commission - this didn’t look like one company simply supporting another.
3:42
It looked like a merger. Naturally, Microsoft and
3:45
OpenAI didn’t label it that way. But the facts were clear to see.
3:49
Microsoft had poured in billions and secured exclusive rights for its Azure ecosystem.
3:54
OpenAI technology was also becoming increasingly integrated in Microsoft’s
3:58
most systems and applications, from Windows to Copilot, Office, GitHub, and beyond.
4:04
OpenAI, meanwhile, was looking less like an independent organization
4:08
and more like a subsidiary, just with its own separate branding.
4:12
The FTC noticed. So did other tech brands.
4:15
Google even called on the government to investigate and break up the deal.
4:19
Critics and regulators argued that Microsoft’s massive investment and exclusive cloud control
4:23
over OpenAI’s technology gave the company too much influence. It hadn’t just invested
4:28
in an up-and-coming company. It had bought the future.
4:32
So, the FTC started investigating the two companies. If it could prove that they had
4:36
effectively entered a de facto merger or that Microsoft had acquired an unfair monopoly over the
4:41
AI industry, it could force the pair to split. Microsoft would be able to survive that.
4:46
OpenAI might not. It couldn’t afford the risk and had to find
4:50
some way to wriggle out of its predicament. Enter the Amazon deal.
4:54
By pivoting to Amazon Web Services (AWS), OpenAI gave itself a multibillion-dollar legal shield.
5:01
Because now, if the FTC’s investigators question the company’s allegiances or argue
5:05
it’s too friendly with Microsoft, OpenAI’s lawyers can simply say:
5:09
“How can we possibly be a subsidiary of Microsoft if we just signed a $50 billion
5:14
contract with one of their biggest rivals?” It was the perfect piece of legal theater.
5:19
At just the right time. Because antitrust cases are built
5:22
on dependency. Regulators are very wary of any company that appears to be entirely or exclusively
5:28
dependent on another for its survival. But by inking an agreement with another tech giant,
5:33
OpenAI proved its “independence.” Problem solved.
5:37
Or, at least, that’s how it seemed. In reality, there was much more to
5:40
this story than meets the eye. Escaping the FTC was only a convenient and timely
5:44
by-product of a deeper and darker machination. The real reason OpenAI needed leverage over
5:50
Microsoft was hidden inside a bizarre legal contract signed years before…
5:55
A contract that contained a ticking time bomb. Chapter 3 - The AGI Poison Pill
6:01
For years, OpenAI told the world that it’s working towards AGI, Artificial General Intelligence.
6:08
Sometimes known as the “god model,” this is said to be the point at which AI effectively reaches,
6:13
and then surpasses, human-level intelligence. According to the experts, AGI will be able to
6:18
think, reason, and adapt like a real person. It will solve problems and switch from task to task,
6:23
rather than being pre-programmed with just one specific function or avenue of activity in mind.
6:28
In effect, this was OpenAI’s justification for everything.
6:32
All the funding, all the hype, all the resources. It was all
6:36
said to be in service of the AGI experiment. And for OpenAI, it was the perfect panacea.
6:42
All they had to do was convince people to trust them, ignore the obvious problems, hand over
6:47
their money, and believe that someday they’d build something that would change the world.
6:51
There was just one little problem. One buried deep in the contract
6:56
tying OpenAI and Microsoft. It was known as the “AGI trigger.”
7:01
Basically, Microsoft was granted a seemingly perpetual license to OpenAI’s intellectual
7:05
property. But that “perpetual license” had a strict limit. As soon as OpenAI achieved
7:11
Artificial General Intelligence, that new AGI model would be entirely excluded from the deal,
7:17
and Microsoft would effectively lose its grip on the future of AI.
7:21
At that stage, all commercial rights to the AGI would remain exclusively with OpenAI
7:27
It’s a paradox. A snake eating its own tail. Microsoft was pouring billions into a company
7:32
whose sole stated mission was “Build AGI.” But as soon as that mission was achieved, Microsoft would
7:38
lose all of its power and benefits. It was funding its own demise.
7:42
But Microsoft’s executives aren’t idiots. They knew the terms of the deal when they signed it.
7:46
They knew exactly how to work around them. All they had to do was ensure that OpenAI
7:51
failed at its stated mission. They wanted the AI to be powerful,
7:55
but never powerful enough to reach AGI. And that’s where things get complicated.
8:00
Because AGI isn’t a clear finish line. No one can agree what it actually means. So even if
8:06
OpenAI pushed its systems further and further, Microsoft could always argue it still wasn’t AGI.
8:12
Sam Altman knew this. He knew that as long as the
8:14
AGI trigger clause existed, Microsoft would never truly be an all-in partner. It would
8:20
always have leverage. Always have limits. Always have a way to control OpenAI from the inside.
8:25
So he changed the play. With the $50 billion Amazon deal as its loaded
8:30
gun, OpenAI forced Microsoft back to the table. Not just to negotiate, but to completely
8:35
rewrite the rules of the AI war. Chapter 4 - The April 2026 Reset
8:41
In April 2026, the balance of power shifted. This wasn’t a simple partnership update.
8:46
OpenAI didn’t want to iron out a few issues or make a couple of amendments to its Microsoft deal.
8:51
It wanted to demolish it, then rebuild it, exactly as it saw fit.
8:56
On the surface, the two companies saved face, announcing a “simplified” agreement and “next
9:00
phase” for their partnership. But the terms of that
9:03
agreement painted the real picture. Microsoft was still described as OpenAI’s “primary
9:08
cloud partner,” but the very next sentence added that OpenAI was now free to serve products across
9:14
any other cloud provider it wanted. Azure’s exclusivity was gone.
9:19
Microsoft’s once “perpetual” license to OpenAI’s intellectual property was also amended and given
9:24
a fixed end date of 2032. The tech giant no longer had privileged ownership of the
9:30
future of AI. Revenue sharing was officially given a cap and deadline of 2030, as well.
9:36
Most importantly, the AGI trigger was gone. It wasn’t redefined, clarified, or amended.
9:42
It was deleted. That vague, hard-to-define
9:44
clause that had hung over OpenAI like a Sword of Damocles for years was gone. It was replaced with
9:50
something far simpler… and far more controlled. A calendar. With dates, deadlines, and caps.
9:56
In other words… a standard corporate agreement. Microsoft’s stake was also formalized at
10:00
approximately 27% of the company. That’s still a sizable amount,
10:04
enough for Microsoft to hold some level of influence over the company’s activities,
10:08
but nowhere near enough for complete control. It looked like a big victory for OpenAI.
10:14
The company had won its independence, decoupling its finances from the mythical AGI milestone. It
10:20
was no longer a research lab trying to trigger a clause in a contract to win its freedom,
10:24
but a corporation with a clear runway ahead. But there was a catch. A big one.
10:30
To pull off this extraordinary corporate coup, Altman had to permanently destroy the very
10:35
foundation that OpenAI was built upon… Its non-profit structure.
10:40
Chapter 5 - The Death of the Non-Profit OpenAI’s non-profit nature was the one
10:45
thing that separated it from every other Silicon Valley machine.
10:49
This wasn’t another power to “the benefit of humanity.”
10:52
A non-profit lab, working with care and consideration towards something that was
10:56
supposed to bring great benefits to all. This wasn’t Google. It wasn’t Meta.
11:00
It wasn’t worried about pleasing its shareholders… because there were no shareholders.
11:05
But that idealistic attitude couldn’t last. Slowly but surely, the cracks in the mask
11:10
began to appear. In late 2025, the façade was ripped away entirely.
11:14
OpenAI shifted from a non-profit to a Public Benefit Corporation.
11:18
At a glance, that still sounds like a righteous cause, a compromise between the original mission
11:23
and the need to make money. PBCs are supposed to strike a balance between pursuing profit while
11:28
remaining committed to creating a positive impact on society, the community, or the environment.
11:34
In reality, PBCs still serve their investors almost as much as any other for-profit entity.
11:39
Just as Altman would go on to smash and rebuild his deal with Microsoft, he also destroyed what
11:44
OpenAI once was, reconstructing it as something completely different.
11:48
The groundwork was laid back in 2023. The organization's original non-profit
11:53
board - the one briefly fired Altman - was removed and replaced by Silicon Valley
11:58
insiders and former Treasury officials. The safety guardrails that had once been
12:02
so critical to the organization’s overall mission were dismantled.
12:06
The systems that had kept OpenAI’s progress in line with its focus on helping humanity
12:11
were gone. In their place were “product safety teams” more concerned with ensuring that their
12:16
AI doesn’t say anything that might offend a big B2B client than harm real people.
12:22
The cogs inside the OpenAI machine were replaced, piece by piece.Until
12:26
something fundamentally changed. What began as a research-driven system started to look like
12:31
something else entirely. A profit engine. One that was preparing for a massive IPO.
12:36
And was increasingly insulated from any meaningful ethical oversight.
12:40
Behind closed doors, Altman and OpenAI’s research leads had realized a terrifying
12:45
technical truth. They weren’t moving towards AGI, as they originally expected.
12:50
They were moving towards a brick wall. Chapter 6 - The Scaling Wall
12:55
For years, the AI industry has relied on an unwavering belief
12:59
in a premise known as “Scaling Laws.” If you add more data and more compute,
13:04
your AI models will become exponentially smarter. It’s all a question of resources. Provide more
13:09
resources and get a better product. All companies like OpenAI had to do
13:13
was keep on building bigger and better data centers. They had to invest in more powerful
13:18
chips and processors. Then they could sit back and watch as their AI followed the
13:22
linear path to “God-like” intelligence. It sounds pretty straightforward.
13:26
And the Scaling Laws worked, for a while.
13:29
Each new generation of AI technology felt like a big leap forward.
13:33
GPT-2 was impressive. GPT-3 was next-level. GPT-4 exceeded expectations.
13:39
So naturally GPT-5, codenamed “Orion,” was expected to be a game-changer. Maybe even the
13:45
final step toward AGI. Or… not.
13:48
Internal reports suggest that the Scaling Laws are stalling. Instead of providing some
13:52
sort of intelligence quantum leap, GPT-5 has hit diminishing returns.
13:57
It’s still getting smarter, but at a slower rate than ever before.
14:00
All of a sudden, the “God model” that seemed right around the corner is now a speck on the horizon.
14:06
This isn’t just a hurdle, it’s a catastrophe. The Scaling Wall changes everything. It’s no
14:11
longer a situation where companies can pour in money and watch their AI become twice as
14:15
intelligent overnight. Now, they’re spending billions for only incremental improvements.
14:21
And that’s bad business. Because let’s not forget
14:24
about the burn rate. OpenAI is nowhere close to making a profit. It loses billions each year.
14:29
But it was always able to justify that with the claim that AGI would eventually
14:33
arrive and fix everything. The company’s entire financial structure and investment
14:37
incentives were reliant on that premise. That’s why removing the AGI trigger clause
14:42
from the Microsoft contract mattered so much. It wasn’t just a legal trick. It was a quiet
14:47
admission that AGI is a mirage. And if AGI is a mirage, OpenAI
14:52
is just another software company… one that is failing to provide returns and plateauing fast.
14:57
So, how does a company like that justify a $5 billion burn rate to its next investors?
15:03
It stops selling AGI and starts selling AI slop instead.
15:07
Chapter 7 - The SaaS Pivot (Selling AI Slop) Meet the new OpenAI.
15:13
It’s no longer a valiant non-profit, pursuing civilization-changing superintelligence, but
15:18
a “Salesforce for AI” business. It’s slowly but surely pivoting away from its original mission and
15:23
towards something much more mundane: enterprise software and corporate workflow automation.
15:28
It’s happening right before our eyes. Rather than focusing its efforts
15:32
exclusively on the next evolution of GPT technology, OpenAI is prioritizing
15:37
alternative projects, like “Agentic” middleware and reasoning models, like o1 (Strawberry).
15:43
It’s no longer charting a course towards human enlightenment,
15:46
but making life easier for middle management. The focus has shifted to producing tools built for
15:51
middle management. Routing leads. Generating marketing copy. Handling support tickets.
15:56
OpenAI hopes that this shift will bring in the money it needs to satisfy its investors.
16:01
But there’s a massive problem with that plan… The numbers don’t add up.
16:05
Traditional SaaS (Software as a Service) companies, like Salesforce and Adobe, operate
16:10
on incredible margins, often exceeding 70%. They make a product and basically sell it forever,
16:16
bringing in more profit with every new customer. That’s why investors love SaaS. It’s a goldmine.
16:23
But OpenAI’s attempts to enter this industry aren’t working.
16:27
Because advanced reasoning models cost so much more to run than conventional software.
16:32
o1, for example, costs around $15 for one million input tokens.
16:36
That might not sound like much at first glance, but in enterprise terms,
16:40
it is a massive financial burden. Big businesses with hundreds or even thousands of employees can
16:45
chew through millions upon millions of tokens in a single day. Suddenly, a “smart assistant” is
16:50
not a cost-effective component of the tech stack, but a very expensive capital drain.
16:55
Traditional SaaS doesn’t work that way. Microsoft doesn’t bill businesses every
17:00
time they open a new spreadsheet on Excel. CRMs don’t suddenly become twice as expensive
17:05
just because employees clicked a few buttons and generated some reports.
17:08
AI works differently. The more you use it, the more expensive it gets.
17:12
OpenAI is desperately trying to force businesses to integrate overpriced,
17:16
automated “AI slop” software into their corporate workflows to fix its own broken economics. It’s
17:22
trying to sell digital gold for the price of lead, hoping to convince people its money-guzzling AI
17:28
agent is just “really expensive software.” But to make its margins work, it needs to
17:33
dramatically decrease its own operating costs. It needs cheaper compute, which brings us back
17:38
to the $50 billion Amazon Trojan Horse. Chapter 8 - Amazon’s Trojan Horse
17:44
The deal with Amazon wasn’t only about escaping the FTC’s investigations or breaking free of
17:49
Microsoft’s shackles. It was about hardware.
17:52
Part of the $50 billion commitment that Amazon made to OpenAI includes a multi-year agreement
17:58
for the AI firm to use AWS’ “Trainium” chips. Prior to this, OpenAI was a hostage to Nvidia.
18:05
It was forced to use Nvidia’s “H100” GPUs. It's the hardware that everyone in the AI
18:10
industry wants and needs. The chips that form the beating hearts of AI data centers.
18:15
These GPUs have proven highly effective in training and improving AI.
18:19
They’re also expensive. Each one can cost tens of
18:22
thousands of dollars. That’s before the added expense of building the rest of the server
18:27
around it and actually running the whole thing. Large training clusters come with billion-dollar
18:31
price tags. OpenAI has paid a “H100 tax” on every single prompt it processes for years,
18:37
with incalculable amounts of money funneled away into the accounts of Nvidia and Microsoft.
18:43
The Amazon deal gives OpenAI an off-ramp. Trainium chips aren’t necessarily better than
18:48
Nvidia’s H100s, but they do have the potential to be much, much cheaper. Up to 50% cheaper,
18:54
according to early estimates. They’re also said to consume 40% less energy.
18:58
By switching to Amazon’s own custom silicon, OpenAI hopes it will be able to make some
19:03
significant reductions to the cost of its tokens. Cheaper tokens should make OpenAI’s AI slop easier
19:09
to digest for its big business customers. They may even give the company a slim chance of turning
19:13
a profit before the 2030 revenue cap hits. In the name of “saving humanity” and building
19:19
a tech utopia, OpenAI took a very different path. Building software on proprietary Amazon
19:24
chips. Running inside Amazon data centers. Selling AI agents to Fortune 500 companies.
19:30
This is not what the company’s founders envisioned all those years ago.
19:34
This is not a beacon of open-source enlightenment. It’s just a cog in the AWS machine.
19:39
So where do we go from here? Chapter 9 - The Great AI Realignment
19:44
The hardware war is over. Unfortunately, humanity didn’t win.
19:48
Instead, the victors are the corporate behemoths that own the silicon.
19:51
The companies with the money and power to do whatever they want and always get away with it.
19:56
OpenAI sold the world a dream. A dream of a god-like AI that would cure cancer,
20:01
solve the climate crisis, and bring about a new world where everyone would be happier,
20:05
freer, and more fulfilled. That utopian dream is dead,
20:09
replaced by a dystopian corporate reality. Microsoft, Amazon, and OpenAI aren’t laying
20:14
the foundations for a more prosperous and creative age of human advancement.
20:18
They’re building their own locked-down and ludicrously expensive B2B monopoly.
20:23
The “open marriage” that now exists between these tech giants all but guarantees that
20:27
the future of the internet will be flooded with corporate AI slop. Automated emails,
20:33
synthetic reports and agentic workflows that don’t actually provide real benefits to real people.
20:38
They just streamline the capitalist machine while eroding the value of human thought and creativity.
20:43
Sam Altman didn’t escape from Microsoft to bring about a better world.
20:46
He broke free so that when the trillion-dollar IPO arrives, he and his shareholders
20:51
will make as much money as possible. This is the grim reality of AI today.
20:56
We’re not getting AGI. We’re not going to see some
20:58
digital God that solves the world’s ills and makes us all happier and healthier.
21:02
We’re getting an inescapable, automated corporate bureaucracy instead.
21:06
And once you follow that logic all the way through, one question becomes unavoidable:
21:11
what happens when the money stops making sense?
21:14
Find out in “$115 Billion Burn Rate. The AI Bubble Just POPPED.” Or watch this instead.
