Latest / Connecting the Dots / Chinese AI Controls, US Adoption Trends, and Alibaba's Monetization Puzzle

Transcript

Chinese AI Controls, US Adoption Trends, and Alibaba's Monetization Puzzle · Connecting the Dots

• 0:00
  Today's tech news moves fast, but understanding
• 0:03
  it shouldn't be hard. That's why Snarfl Solutions
• 0:06
  Group brings you Connecting the Dots, where we
• 0:09
  break down what's happening in tech and why it
• 0:11
  matters to you. Now, let's dive in with your
• 0:14
  hosts. Imagine running a business where the core
• 0:20
  engine powering your entire company. You know,
• 0:23
  your customer service, your data analysis, your
• 0:25
  internal coding is just suddenly outlawed overnight.
• 0:28
  Yeah, that would be terrifying. Right. Today,
• 0:30
  American companies are quietly migrating to Chinese
• 0:33
  artificial intelligence to save, I mean, literally
• 0:36
  millions of dollars. But they might actually
• 0:40
  be building their futures on a massive geopolitical
• 0:42
  fault line. Welcome to Connecting the Dots, the
• 0:45
  daily show where we break down the latest tech
• 0:46
  news and figure out exactly how it impacts your
• 0:49
  everyday life. I'm your host, Alex. And I'm Morgan.
• 0:51
  We're coming to you live from Starful Solutions
• 0:53
  Group right here in Sacramento, California. And
• 0:56
  it is Tuesday, July 7, 2026. And we have a really
• 0:59
  packed show lined up for you today. We really
• 1:01
  do. But before we get into the heavy hitting
• 1:04
  tech news and that massive AI shift, let's get
• 1:06
  you caught up on the daily briefing so you are
• 1:09
  fully prepared for the day ahead. Starting off
• 1:11
  with the weather right here in Sacramento is
• 1:13
  a beautiful sunny Tuesday. We were looking at
• 1:16
  a high of 87 degrees Fahrenheit with a nice light
• 1:19
  wind of 512 miles per hour. Absolutely perfect
• 1:23
  summer conditions. It really is gorgeous locally,
• 1:26
  though we should note for everyone listening
• 1:27
  across the rest of the country, things are significantly
• 1:30
  more turbulent outside of California right now.
• 1:32
  As of this morning, there are currently 195 active
• 1:36
  U .S. weather alerts, with 84 of those situated
• 1:39
  over land. So depending on where you are tuning
• 1:41
  in from today, make sure you double check your
• 1:43
  local forecast before heading out the door. Good
• 1:45
  call. Always better to be prepared. Now, pivoting
• 1:48
  over to the markets, if you were watching the
• 1:49
  tickers yesterday, Monday brought some very solid
• 1:52
  green across the board. The Dow Jones jumped
• 1:54
  up over 227 points to close out at 53 ,055 .91.
• 1:59
  Yeah, and the S &P 500 mirrored that positive
• 2:02
  momentum as well, rising 30 points to finish
• 2:05
  the day at 7 ,537 .43. On the crypto side of
• 2:10
  things, though, we are seeing a slight pullback
• 2:12
  this morning. As of today, Bitcoin dipped 1 .5%,
• 2:16
  bringing it down to just over 63 ,000. A little
• 2:19
  bit of a breather for the crypto markets, it
• 2:21
  seems. Finally, let's wrap up the morning briefing
• 2:24
  with a little bit of sports. A massive congratulations
• 2:27
  to the San Francisco Giants, who absolutely crushed
• 2:30
  the Toronto Blue Jays yesterday with a final
• 2:32
  score of 10 to 1. There wasn't even a contest,
• 2:34
  honestly. Landon Rupp was the winning pitcher,
• 2:36
  and he threw an absolute gem of a game. A thoroughly
• 2:39
  great night for baseball fans on the West Coast.
• 2:41
  Yeah, seriously, they totally dominated. All
• 2:43
  right, with the morning updates out of the way,
• 2:45
  we are ready to get into today's Morning Star
• 2:47
  Tech Report, where the global landscape of artificial
• 2:49
  intelligence is just, well, fracturing in real
• 2:51
  time. But first, a quick word from our sponsor.
• 2:54
  Today's episode is brought to you by Strong Coffee,
• 2:56
  the premium instant coffee that tastes like a
• 2:58
  traditional brew but is packed with 15 grams
• 3:00
  of protein, MCTs and adaptogens like ashwagandha.
• 3:04
  Get clean, sustained energy without the jitters.
• 3:07
  Use promo code SNARFL for 20 % off at strongcoffeecompany
• 3:11
  .com. Okay, let's unpack this. We really need
• 3:14
  to talk about what is quietly becoming the biggest
• 3:17
  and potentially the most disruptive shift in
• 3:19
  enterprise tech this year. We're witnessing what
• 3:22
  I'm calling the great AI migration. U .S. companies,
• 3:26
  meaning the businesses you interact with, buy
• 3:28
  from and work for every single day, are actively
• 3:31
  abandoning American -built AI systems in droves.
• 3:33
  Yes, wild to watch. It is. Instead, they're adopting
• 3:36
  Chinese -built AI models. to run their daily
• 3:39
  operations. The sheer scale of the shift is really
• 3:41
  what makes this so critical to understand. If
• 3:43
  we look at the data coming out of OpenRatter,
• 3:45
  which is essentially a massive marketplace platform
• 3:48
  that developers use to plug different AI models
• 3:50
  into their software, U .S. companies have directed
• 3:52
  over 30 % of their token use to Chinese models
• 3:55
  every single week since February 8th of this
• 3:58
  year. Wow, that is a huge chunk of the market.
• 4:01
  Right. And at one specific point, it actually
• 4:03
  peaked at 46%. Let me pause you there. Because
• 4:06
  we throw the word token around a lot in tech,
• 4:09
  and I want to make sure everyone listening is
• 4:10
  on the exact same page. Think of a token as the
• 4:13
  raw currency of artificial intelligence. It's
• 4:16
  roughly equivalent to a word or maybe a syllable.
• 4:20
  Every single time you ask an AI a question or
• 4:23
  generates an answer for you, you are spending
• 4:25
  tokens. That is a great way to explain it. Thanks.
• 4:29
  So when we say 30 % of token use, we mean 30
• 4:32
  % of that word currency, the actual thinking
• 4:34
  and processing power, is now flowing directly
• 4:36
  into Chinese models. Exactly. And to give you
• 4:39
  some historical context on that 30 to 46 % figure,
• 4:43
  prior to this recent surge, Chinese models only
• 4:45
  accounted for about 11 % of that token trapping
• 4:48
  over the previous 12 months. If you go back just
• 4:50
  a little further to early 2025, it was sitting
• 4:53
  at a mere 4 .5%. So we have gone from practically
• 4:56
  zero market share to almost half the pie in certain
• 5:00
  developer ecosystems in just a matter of months.
• 5:03
  Which naturally brings up the biggest question.
• 5:05
  Why? Why are American businesses, who have access
• 5:09
  to domestic giants like OpenAI and Anthropic,
• 5:12
  suddenly pivoting to foreign infrastructure for
• 5:15
  their core technology? Honestly, it comes down
• 5:17
  to basic economics. I mean, the cost of running
• 5:19
  these advanced, bleeding -edge American AI systems
• 5:22
  has just skyrocketed. When you are a large enterprise
• 5:26
  trying to integrate AI into every single department,
• 5:29
  automating customer service chats, to analyzing
• 5:32
  massive financial data sets, to generating thousands
• 5:34
  of lines of code, you are burning through billions
• 5:37
  of tokens. Oh, absolutely. And those costs scale
• 5:39
  up incredibly fast, often hitting millions of
• 5:42
  dollars a year just in API fees alone. Exactly.
• 5:44
  So it is entirely a budget play. Chinese models
• 5:47
  have recently hit the market, offering prices
• 5:49
  that are 60 to 90 percent cheaper than the top
• 5:52
  American systems. And the crucial piece of the
• 5:54
  puzzle here is how they are doing it. It isn't
• 5:57
  just about heavily subsidized compute or cheap
• 5:59
  labor, which is what a lot of people assume.
• 6:01
  Right. There is actual technological innovation
• 6:03
  happening there. Yes. Chinese AI labs have been
• 6:07
  forced to become incredibly efficient. They are
• 6:10
  utilizing highly optimized smaller architectures,
• 6:13
  like a framework called mixture of experts. Without
• 6:17
  getting too deep into the weeds, instead of firing
• 6:19
  up the entire massive brain of the AI for every
• 6:22
  single question, this architecture only activates
• 6:25
  the specific small part of the brain needed to
• 6:29
  answer that exact prompt. Ah, so it requires
• 6:32
  significantly less computing power, which drops
• 6:35
  the operating cost through the floor. You know,
• 6:37
  it reminds me heavily of the pharmaceutical industry.
• 6:39
  Think about when a doctor writes you a prescription.
• 6:42
  You can go to the pharmacy and buy the shiny
• 6:44
  name -brand medication for an absolute fortune,
• 6:46
  or you can buy the high -quality generic version.
• 6:48
  Oh, that's a perfect analogy. Right. If that
• 6:51
  generic pill is 90 % as effective but costs a
• 6:53
  fraction of the price, budget -conscious businesses
• 6:56
  have no choice but to make the switch. They're
• 6:58
  simply acting rationally to protect their bottom
• 7:00
  line and keep their own products affordable for
• 7:03
  you, the consumer. That generic drug analogy
• 7:05
  is spot on. And the industry analysts completely
• 7:08
  back this up. They estimate that these cheaper
• 7:11
  Chinese models are only trailing about six to
• 7:14
  nine months behind the absolute frontier capabilities
• 7:17
  of the best American systems, which is. Really
• 7:20
  nothing the grand scheme of things. Exactly.
• 7:21
  For 95 % of business tasks, you really don't
• 7:24
  need the model that was invented yesterday. You
• 7:26
  just need the one from six months ago that costs
• 7:28
  pennies on the dollar. And we're seeing this
• 7:30
  play out in real time with some staggering specific
• 7:32
  examples. Look at a Beijing based AI startup
• 7:36
  called Z .AI. They recently released a model
• 7:39
  named GLM 5 .2. When it hit the Vercel platform,
• 7:43
  which is a major hub developers use to build
• 7:46
  and deploy web applications, its daily token
• 7:49
  volume grew 27 times over in just its first week.
• 7:52
  That is massive growth. Going from a handful
• 7:55
  of beta testers to multiplying your volume by
• 7:58
  27 in seven days tells you this isn't just a
• 8:00
  shiny doi developers are testing out for fun.
• 8:03
  Businesses are actively ripping out American
• 8:05
  infrastructure and replacing it. Yeah, they really
• 8:07
  are. And the reason is that GLM 5 .2 is performing
• 8:10
  almost identically on key benchmarks to Anthropic's
• 8:15
  highly advanced Opus 4 .8 model, but it is doing
• 8:17
  it for literally one fifth of the price. Then
• 8:20
  you have an AI startup like Lindy. They had built
• 8:23
  their operations using Anthrobic's Claude model,
• 8:25
  but in June, they moved 100 % of their traffic
• 8:28
  over to a Chinese model made by DeepSeek. The
• 8:31
  CEO of Lindy literally came out and said their
• 8:34
  cost curve just crashed to the ground. They are
• 8:38
  saving millions of dollars simply by swapping
• 8:40
  out the underlying engine. The margins are just
• 8:42
  way too good for tech leaders to ignore. But
• 8:44
  wait, if I'm looking at this from Beijing's perspective,
• 8:47
  I am absolutely thrilled, right? I mean, Western
• 8:50
  companies are willingly handing me massive market
• 8:52
  share. They're relying on my infrastructure and
• 8:54
  they're paying me for it. Why would I want to
• 8:57
  stop that momentum? Yeah, that is exactly where
• 8:59
  the narrative takes a sharp turn. Let's look
• 9:01
  closer at the reaction back in China, because
• 9:03
  as these Chinese models become globally dominant
• 9:05
  and deeply embedded into foreign corporate infrastructure,
• 9:09
  Beijing is suddenly getting very nervous. Really?
• 9:12
  nervous about winning the market. Yeah, they
• 9:14
  are realizing they might be giving away their
• 9:16
  most valuable strategic technological assets
• 9:19
  for pennies. Oh, wow. So they capture the global
• 9:22
  market and now they are considering locking the
• 9:24
  gates. Precisely. China's Ministry of Commerce
• 9:27
  has actually spent the last month convening high
• 9:30
  level meetings with major tech giants like Alibaba,
• 9:33
  ByteDance and that startup we just mentioned,
• 9:36
  ZAI. OK, so all the big players. All of them.
• 9:38
  And the central topic of conversation is how
• 9:40
  to restrict overseas access to their top tier
• 9:43
  AI models. They are discussing potential regulatory
• 9:47
  curbs on both their closed source systems which
• 9:50
  are accessed via APIs and their open weight models.
• 9:54
  I have to stop you there because the logistics
• 9:56
  of this make absolutely zero sense to me. I understand
• 9:59
  cutting off API access, you just turn off the
• 10:01
  server. But how on earth do you put the genie
• 10:04
  back in the bottle when it comes to open weight
• 10:06
  models? It's a great question and it's complicated.
• 10:08
  Right, because for our listeners, open wait basically
• 10:11
  means the core instructions, the actual brain
• 10:14
  of the AI, have already been packaged up and
• 10:16
  released for developers to download and run locally
• 10:19
  on their own physical computers. Once it's downloaded
• 10:21
  by a developer in Ohio, you can't just recall
• 10:24
  it. It's like trying to unsend a mass email.
• 10:26
  You are entirely correct about the models that
• 10:28
  are already out in the wild. You cannot wipe
• 10:30
  a hard drive in Ohio from a desk in Beijing.
• 10:33
  But from a regulatory standpoint, they aren't
• 10:36
  trying to delete history, they're trying to control
• 10:39
  the future. The measures they are proposing are
• 10:42
  severe. Like what kind of measures? Well, they
• 10:44
  are considering classifying the unauthorized
• 10:47
  leak or overseas release of proprietary AI as
• 10:50
  a direct violation of their national security
• 10:53
  law. Ah, I see. So if a Chinese developer knows
• 10:56
  that putting the next iteration of their model
• 10:58
  on the open internet could result in major national
• 11:00
  security charges, they simply won't do it. Exactly.
• 11:03
  It creates a massive chilling effect on the entire
• 11:06
  industry. They are also talking about limiting
• 11:08
  who can invest in domestic AI startups specifically
• 11:12
  to keep foreign capital and foreign influence
• 11:14
  out of their ecosystem. That makes sense from
• 11:17
  a protectionist standpoint. Yeah, and there is
• 11:19
  also a proposal on the table to sort all AI development
• 11:22
  by sensitivity levels. Routine basic tools might
• 11:25
  just need standard registration, but the most
• 11:28
  powerful frontier systems would be legally mandated
• 11:31
  to remain entirely domestic. No public or international
• 11:35
  release whatsoever. Okay, so Beijing is terrified
• 11:38
  of their top tier AI leaking out and fueling
• 11:41
  foreign companies. But that paranoia didn't just
• 11:43
  appear out of nowhere. If you look at the timeline,
• 11:46
  they're essentially holding up a mirror to Washington.
• 11:49
  This isn't happening in a vacuum. It's a direct
• 11:50
  response to recent moves by the United States.
• 11:53
  It really is a massive geopolitical AI tit -for
• 11:56
  -tat. And we should be very clear to you, our
• 11:58
  listener, we are strictly reporting on the policy
• 12:01
  moves and the stated anxieties of both the US
• 12:03
  and Chinese governments here. We aren't taking
• 12:06
  a side or endorsing either nation's viewpoint.
• 12:09
  Exactly. We are just laying out the mechanics
• 12:12
  of how this global standoff is escalating because
• 12:14
  it directly impacts the software you use every
• 12:17
  day. So let's look at what the U .S. has been
• 12:20
  doing that triggered this whole reaction. Recently,
• 12:23
  the U .S. government stepped in and forced Anthropic
• 12:26
  to temporarily ban foreign access to its most
• 12:29
  advanced models, specifically the ones named
• 12:32
  Mythos and Fable. Right, and they cited strict
• 12:34
  national security concerns for that move. They
• 12:37
  did. Now, they did eventually ease some of those
• 12:40
  restrictions on the Mythos model, but only for
• 12:42
  select, highly vetted Fortune 500 companies.
• 12:46
  The broader message to the global market was
• 12:48
  sent loud and clear. Yeah, and OpenAI has faced
• 12:51
  very similar pressure from the US government
• 12:53
  behind the scenes, resulting in them holding
• 12:55
  back or delaying the release of certain advanced
• 12:57
  systems. The United States is very clearly treating
• 13:00
  high -level artificial intelligence not just
• 13:01
  as software, but as a strategic national asset
• 13:04
  that needs to be protected with export controls.
• 13:07
  It's very much in the same way they treat military
• 13:09
  technology or advanced semiconductor manufacturing.
• 13:12
  And from the Chinese perspective, this US posturing
• 13:15
  is highly alarming. Officials in Beijing, including
• 13:18
  Zhu Hongyi, who is the founder of the major Chinese
• 13:21
  cybersecurity company 360, are pointing specifically
• 13:24
  at Anthropic's Mythos model. Yeah, Mythos is
• 13:27
  a huge sticking point for them. It is, because
• 13:30
  Mythos is incredibly capable when it comes to
• 13:33
  complex cybersecurity tasks. It can write secure
• 13:36
  code, but the fear in Beijing is the dual -use
• 13:39
  nature of that ability. They worry that a tool
• 13:41
  that smart could be weaponized by the US to aggressively
• 13:44
  scan, identify, and exploit zero -day vulnerabilities
• 13:47
  in Chinese software systems. It is the textbook
• 13:50
  definition of a security dilemma. The US restricts
• 13:53
  its AI because it fears what foreign adversaries
• 13:55
  might do with it. China looks at those restrictions,
• 13:58
  views the American AI as a potential cyber weapon,
• 14:01
  and decides it must hoard its own AI advancements
• 14:03
  to build up comparable defensive and potentially
• 14:06
  offensive capabilities. So neither side trusts
• 14:08
  the other, so both sides just lock down. Exactly.
• 14:11
  And Beijing has already started retaliating in
• 14:14
  other, very tangible ways beyond just proposing
• 14:16
  these new export limits. Right. Didn't they block
• 14:18
  a major acquisition recently? They did. Earlier
• 14:21
  this year, they essentially forced Metta to completely
• 14:24
  reverse a $2 billion deal to acquire a Chinese
• 14:28
  AI startup called Manus. That deal was practically
• 14:31
  done and Regulator stepped in and tore it up.
• 14:33
  They are also actively opening inquiries into
• 14:36
  Chinese AI firms that have relocated their headquarters
• 14:39
  overseas just to see if those companies are trying
• 14:41
  to dodge these upcoming export rules. The walls
• 14:44
  are going up so rapidly on both sides. And this
• 14:47
  brings us to the broader implication for working
• 14:49
  professionals. For you, the listener. We have
• 14:52
  spent the last two decades getting used to a
• 14:54
  totally borderless digital world. You download
• 14:57
  a tool from a server in Europe, you write code
• 14:59
  in California, you deploy a business globally.
• 15:01
  It's just how the internet works, or worked.
• 15:03
  Exactly. But as fundamental technology becomes
• 15:06
  inextricably linked to national security, governments
• 15:09
  are building incredibly high digital walls that
• 15:12
  fracture that ecosystem. Which brings us right
• 15:15
  back to the great AI migration we started with.
• 15:18
  American companies right now are building their
• 15:20
  products. their daily workflows and their entire
• 15:23
  financial cost structures on top of these incredibly
• 15:26
  cheap, highly capable Chinese models. Right.
• 15:29
  They are saving millions today, but they are
• 15:31
  building their houses on a geopolitical fault
• 15:34
  line that could crap at literally any moment.
• 15:37
  It is an incredibly precarious position to be
• 15:40
  in as a business owner. The global tech ecosystem
• 15:43
  is fracturing into regional, heavily regulated
• 15:46
  silos. The startups and enterprise businesses
• 15:48
  caught in the middle are going to have to navigate
• 15:50
  some incredibly complex legal and geopolitical
• 15:54
  minefields just to keep their basic software
• 15:56
  engines running. It really makes you rethink
• 15:58
  the hidden long -term costs of cheap technology.
• 16:02
  We're gonna wrap things up in just a moment and
• 16:03
  leave you with one final thought on all of this,
• 16:05
  but before we do, we have a quick message. Before
• 16:07
  we wrap up, a quick thanks to Old Glory, an iconic
• 16:10
  music and sports fan merch store. Perfect for
• 16:13
  the techie in your life. Use promo code SNARFL
• 16:16
  for 15 % off at oldglory .com. So let's connect
• 16:20
  the dots on everything we've explored today.
• 16:23
  We started by looking at how sky -high, enterprise
• 16:25
  -level computing costs here in the US have essentially
• 16:28
  forced budget -conscious businesses to adopt
• 16:31
  highly optimized, low -cost Chinese AI models
• 16:34
  just to stay financially competitive. Then we
• 16:37
  saw how that exact global success, the massive
• 16:40
  adoption of their developer tokens by Western
• 16:42
  companies, has actually panicked Beijing. It
• 16:44
  is leading them to propose strict export limits
• 16:47
  and national security laws to protect their digital
• 16:49
  crown jewels from being utilized by foreign entities.
• 16:52
  And finally, we saw how all of this is deeply
• 16:54
  tied to an escalating high -stakes cybersecurity
• 16:57
  standoff between two global superpowers. Both
• 17:00
  sides are throwing up regulations, reversing
• 17:02
  multi -billion dollar acquisitions, and treating
• 17:04
  artificial intelligence like the new digital
• 17:06
  nuclear deterrent. It is a rapidly shifting landscape.
• 17:10
  What is absolutely clear is that the era of completely
• 17:13
  open, borderless AI development is facing severe
• 17:17
  geopolitical headwinds that will change how software
• 17:20
  is built moving forward. Which leaves us with
• 17:23
  a pretty intense question for you to ponder today
• 17:25
  as you go about your work. Yeah. If sudden geopolitical
• 17:29
  regulations cut off access to these cheap foreign
• 17:31
  models overnight, it's not just a matter of switching
• 17:34
  back to American AI. Exactly. It begs the question,
• 17:38
  have we reached a point where American -made
• 17:40
  technology is becoming a luxury item that only
• 17:42
  the massive tech giants can afford, leaving everyday
• 17:45
  startups and small businesses totally reliant
• 17:48
  on foreign infrastructure just to survive? I
• 17:50
  mean, if the generic medication suddenly disappears
• 17:52
  from the pharmacy shelves and the name brand
• 17:55
  is too expensive to buy, what happens to the
• 17:56
  patient? It's a reality every tech leader needs
• 17:59
  to be preparing for right now. We're here to
• 18:01
  help at Snarful, so reach out on our website
• 18:03
  if you have any more questions. That's a wrap
• 18:08
  on today's top stories. Remember, tech doesn't
• 18:11
  have to be complicated, and we're here to help
• 18:13
  connect the dots. If you've got questions or
• 18:16
  want to learn more, visit us at snarful .com.
• 18:19
  Thanks for listening, and we'll catch you next
• 18:21
  time.