Kind: captions
Language: en
While you've been watching Open AI and
Google burn through billions on chips
and data centers, a Chinese company
you've never heard of just captured 18%
of Africa's AI market for free.
Microsoft just warned investors that
we've been watching the wrong race
entirely. China isn't trying to outspend
us. They figured out how to win by
making AI so cheap that half the world
can't afford not to use it. And
honestly, it's working.
So, in this video, I'm breaking down the
strategy that has Silicon Valley
scrambling, and it's not what you think.
We're talking about how a company you've
never heard of is beating Chat GPT in
Africa, why Apple just admitted defeat
and outsourced Siri brain to Google, and
what it means when AI agents stop
talking and start actually doing your
work. First up, let's talk about the
warning Microsoft just sent to every
investor in the West. The Microsoft
warning. Microsoft's president Brad
Smith didn't mince words. He said China
is effectively winning the AI race
outside of the West. And here's the
kicker. They're not doing it with better
technology.
They're doing it by making AI free and
open. While American companies are
charging $20 a month for ChatGpt Plus, a
Chinese lab called Deep Seek is flooding
developing countries with high
performance AI that costs nothing.
We're talking about Ethiopia, Zimbabwe,
Bellarus, Cuba, places where $20 might
be someone's daily wage. In these
markets, Deepseek isn't just
competitive, it's dominant. The numbers
are striking.
In some African nations, Deepseek has
captured nearly 20% of the AI market in
less than a year. And they're doing this
because their models come pre-installed
on Huawei phones. No subscription, no
credit card. just open the app and start
using it. This is what Microsoft calls a
geopolitical instrument. China isn't
just selling software. They're building
the digital infrastructure for the next
billion internet users. And once those
users are trained on Chinese AI tools,
guess which ecosystem they're locked
into?
The Deep Seek disruption.
Let's talk about what makes Deepseek so
dangerous to the Silicon Valley model.
For years, we've been sold on the idea
that AI requires massive scale.
You need a hundred billion dollars in
chips. You need data centers the size of
small cities. You need to burn through
energy like there's no tomorrow.
That's been the narrative. Deepseek
proved that narrative wrong. They built
models that match or beat GPT4 while
spending a fraction of what OpenAI
spent. How? They focused on efficiency
instead of brute force. better
algorithms, smarter training techniques,
highquality synthetic data instead of
scraping the entire internet. And here's
where it gets interesting. Because of US
export controls, China couldn't get
their hands on Nvidia's best chips. So
instead of competing on hardware, they
optimized for software. They turned a
restriction into an advantage.
In late 2024, Deepseek released their R1
model under an open-source license. That
means anyone, anywhere can download it,
modify it, and build on top of it. No
licensing fees, no corporate
gatekeepers, just pure access. This move
sent shock waves through Wall Street.
Investors had been betting that
companies like Nvidia would see endless
demand for their chips. But if you can
build worldclass AI with less compute,
suddenly that $600 billion
infrastructure buildout doesn't look
like such a sure thing anymore. China's
Six Tigers, Deepseek, isn't alone.
There's an entire ecosystem of Chinese
AI startups that most people in the West
have never heard of. They're called the
Six Tigers. Jepu AI, Moonshot AI,
Miniaax, 01 AI, Bichuan Intelligence,
and Stepfund. These companies have
collectively raised billions of dollars,
mostly from Chinese tech giants like
Alibaba, Tencent, and Shyomi. And
they're not just copying Western models.
They're innovating in ways that make
Western companies nervous. Take Zepoo
AI. They just filed for an IPO in Hong
Kong to become the world's first
publicly traded large model stock. Their
GLM4 Plus model competes directly with
GPT4. And their voice model can hold
real-time conversations in Chinese and
English with almost no latency. Or look
at Moonshot AI. Their Kimmy chatbot can
process 2 million characters at once.
That's the equivalent of reading five
entire books and answering questions
about all of them simultaneously.
For anyone doing research or dealing
with massive documents, that's a
gamecher.
Then there's 01 AI founded by Kyu Lee,
one of the most respected figures in
global tech. Their Ye Lightning model
topped the LMSYS benchmark while being
trained on a compute budget that would
make OpenAI laugh.
They're proving that you can do more
with less and that's terrifying for
companies whose entire business model
depends on spending more. What's
fascinating here is the strategy. While
Western companies compete by building
moes and keeping their technology
proprietary, Chinese companies are
competing by lowering costs and
expanding access. It's a completely
different philosophy and it's working.
Apple's billiond dollar admission. Now,
let's flip to the western side of this
story because the moves happening here
are just as revealing. Apple, the most
valuable company on the planet, just
made a decision that tells you
everything about how fast this race is
moving. They chose Google's Gemini to
power the next generation of Siri and
Apple intelligence. Not their own
models, not some scrappy startup.
Google.
Think about what that means. Apple has
spent years building the narrative that
they do everything in house, that they
control the full stack, that privacy and
integration are their differentiators.
And then they looked at the AI landscape
and said, "We can't do this alone."
According to reports, Apple is paying
Google roughly a billion dollars a year
for this partnership. That's how much it
costs to stay relevant in AI right now.
Even with Apple's resources, more than
$3 trillion in market cap, they couldn't
keep pace with the speed of AI
development. But here's the smart part.
Apple isn't putting all their eggs in
one basket. They're using what they call
a multi-provider strategy. For simple
tasks that can run on your device, like
setting a timer or sending a text, Siri
uses Apple's own models.
For general knowledge questions, it taps
Google Gemini. And for complex deep
research tasks, it can route to OpenAI's
chat GPT.
Your phone is becoming a brain router.
It's not running one AI. It's picking
the best AI for each job. That's the
future. And the market noticed. When
this partnership became official,
Alphabet surpassed Apple in market
capitalization. Investors are betting
that the brain is now more valuable than
the device.
If AI is the operating system of the
future, then controlling the
intelligence layer matters more than
controlling the hardware.
The Google monopoly ruling. But Google's
dominance comes with a catch. In August
2024, a US federal judge ruled that
Google holds an illegal monopoly in
online search. The court found that
Google was paying over $20 billion a
year to Apple and Samsung just to remain
the default search engine on their
devices.
That's an insane number. Think about it.
Google was spending $20 billion annually
just to make sure you didn't have to
think about which search engine to use.
And the court said, "That's
anti-competitive.
You're blocking rivals from even getting
a chance." So, what's the remedy? The
court has ordered Google to share parts
of its search index and user interaction
data with competitors like Open AI.
Imagine Google having to hand over the
data that makes its search engine work
to the company that's trying to replace
it. Google is appealing, of course.
They're arguing that forcing them to
share data would risk user privacy and
stifle innovation, but the legal
proceedings in 2025 suggest that the era
of Google's unchallenged search
dominance is ending.
And this is where things get really
interesting because Google knows they
can't rely on traditional search
forever.
That's why they're making aggressive
moves into what they're calling agentic
commerce
from chat bots to action engines. Let's
talk about the shift that's happening
right now. For the past 2 years, AI has
been about chat bots. You type a
question, the AI gives you an answer.
It's helpful, but it's limited. You're
still doing most of the work. Now, we're
entering the age of AI agents. Systems
that don't just answer questions, but
actually complete tasks for you.
This is the difference between asking
for a recipe and having a meal delivered
to your door. Meta just made this shift
concrete by acquiring a startup called
Manis AI for over $2 billion.
Manis isn't a chatbot. It's what they
call an action engine. You can give it a
highle goal like create a presentation
for my pitch meeting and it will
research the topic, write the content,
design the slides, and deliver a
finished PowerPoint in under 30 minutes.
It doesn't wait for you to prompt it
step by step. It just figures out what
needs to happen and does it. That's the
future Meta is betting on. Billions of
users with autonomous assistance
embedded into WhatsApp and Instagram.
Meanwhile, Anthropic, the company behind
Claude, released something called Claude
Co-work.
It's an AI agent that works inside your
Mac. You point it at a folder full of
messy files, receipts, invoices, random
screenshots, and tell it, "Organize this
and make me a tax spreadsheet, and it
just does it. It opens the apps, reads
the data, moves things around, and spits
out a clean result." This is
fundamentally different from what we've
had before. It's not giving you advice,
it's doing your chores.
Google's commerce protocol.
Google sees this shift coming and
they're making a massive play to stay
relevant. They just launched the
Universal Commerce Protocol, an open
standard that lets AI agents handle
shopping on your behalf. Here's how it
works. Instead of you searching for best
running shoes, comparing prices across
five websites, adding items to carts,
entering your payment info, and tracking
shipments, the AI does all of that for
you. You just tell it what you want and
it completes the entire transaction.
Google has partnered with Walmart,
Target, and Shopify to make this happen.
Retailers are hosting what they call
business agents directly in search
results. These agents can answer product
questions, check inventory, and process
orders without you ever leaving Google.
And here's the clever part. Google is
building in Agentic Checkout with price
tracking.
You can tell the AI, "I want this
laptop, but only if it drops below
$800."
And the agent will monitor prices across
the web and automatically buy it when
the threshold is hit. This is Google's
defensive mode. They know that
traditional search traffic is declining
as people move to AI interfaces. So,
they're making sure that even if you're
not searching, you're still transacting
through Google's infrastructure.
Robots learning from YouTube.
The final piece of this puzzle is
robotics. For years, the bottleneck in
humanoid robots has been the lack of
training data. You can't teach a robot
to fold a shirt by typing instructions.
You need millions of examples of
physical movements. Researchers just
solved this problem.
There's a new framework called VPRA,
video prediction for robot actions.
It lets robots learn by watching YouTube
videos of humans doing tasks. The robot
doesn't just memorize movements. It
builds what's called a world model. It
learns the physics of how objects
interact,
how gravity works, how force transfers
when you push something.
It predicts what happens next in a scene
and then figures out what actions would
create that outcome.
This is massive. Companies like 1X are
using this approach to train their
humanoid robots to handle tasks in real
environments without any manual
programming. The robot watches a video
of a human opening a jar, understands
the underlying physics, and then applies
that knowledge to opening any jar, even
ones it's never seen before. We're
moving from robots that follow scripts
to robots that understand the world. And
they're learning from the same videos we
watch on YouTube.
what this all means. So, here's where we
are. China is winning on efficiency and
global distribution. They're capturing
the next billion users by making AI
cheap, open, and accessible. The US is
winning on deep integration and premium
interfaces.
They're embedding intelligence into
devices and services that billions of
people already use.
But the real shift isn't about
geography. It's about what AI is
becoming.
We're moving from systems that talk to
systems that act. From chatbots to
agents, from answering questions to
completing tasks.
Whether it's Siri powered by Gemini,
Claude organizing your files, manis
building your presentations, or robots
learning from YouTube, the goal is the
same. AI that stops explaining and
starts doing. The question isn't whether
this future is coming. It's here. The
question is which ecosystem you'll be
working in and whether you're ready to
let an AI agent handle your shopping,
your taxes, and maybe even your job.
Thanks for watching.
Drop a comment and let me know. Are you
excited about AI agents or does this
feel like we're handing over too much
control?
I'll see you in the next one.