In partnership with
The GTM bets that shouldn't have worked, and did
One grew revenue 50x after half his team quit over the strategy. One brought in 50K signups in a single day with no paid budget. One generated 100M+ views from a stunt that took 50 hours to conceive. One asked every prospect to demo the product themselves instead of demoing it for them.
None of them followed the safe playbook. They treated GTM like an experiment, moved before they had proof, and made bets most founders would never get approved.
HubSpot for Startups documented all 6 stories in the free Bold Bets Playbook. The risks they took, why it was risky, and what it returned.
AI & Technology Intelligence
AI NEWS INSIDER
Issue #62 · June 8, 2026
THIS WEEK: The most build-it-in-house company on earth is renting its AI brain from a rival. Apple's WWDC reveal is a build-vs-buy lesson for every team.
6 MIN READ | Sharp, actionable intelligence
Even Apple Buys: What the Gemini-Siri Deal Teaches Every Team About AI
By AI News Insider Editorial · 5 min read
Apple builds its own chips, its own operating systems, its own maps, its own modems. The company's entire identity is "we make it ourselves." So the headline from today's WWDC keynote should stop you cold: Apple is rebuilding Siri on Google's Gemini, not its own models. The most vertically integrated company on earth chose to buy the AI brain it could not build fast enough.
The deal, and why the math is brutal
Apple is using today's keynote to show off a Siri rebuilt on a custom 1.2-trillion-parameter Gemini model licensed from Google, a deal Bloomberg's Mark Gurman pegs at roughly $1 billion a year. The new Siri arrives as a standalone app with a chatbot interface, a system-wide "Search or Ask" gesture, and a spot in the Dynamic Island. It gets personal context too: access to your emails, photos, and files, processed through Apple's own infrastructure. This is Tim Cook's final WWDC keynote before John Ternus takes over, staged under the tagline "All Systems Glow."
Read the decision plainly. Apple spent years and a reported fortune trying to ship a next-generation Siri in-house, and it slipped twice. The honest internal verdict was that matching frontier quality on its own timeline was not realistic. So Apple did the unsentimental thing: it bought the capability and redirected its own talent to the parts only Apple can do.
The part most coverage misses: Apple kept control
Buying the model is not the whole story. Apple routed Gemini through its own Private Cloud Compute, so the privacy promise and the user relationship stay with Apple, not Google. And iOS 27 Extensions let users swap in Claude or Gemini as their assistant, with an AI agent tier reportedly coming to the App Store. Apple bought the engine but kept the steering wheel. If Google's model stops being the best option, Apple can change the part without rebuilding the car.
That is the difference between a smart buy and a dependent one. The losing version of "buy" hands a rival your roadmap. Apple's version of "buy" treats the model as a swappable component behind an interface it controls.
Why this lands on your desk, not just Apple's
Every team building with AI right now faces the same fork Apple just took. Train or fine-tune your own model, or call someone else's API. Most companies pour money into building capabilities that are not their actual product, then watch a frontier lab leapfrog them in a quarter. If Apple, with its cash and talent, concluded that building the model was the wrong fight, the bar for your team to build instead of buy just got a lot higher.
The Bottom Line for AI News Insider Readers
Even Apple buys when the capability is not its edge. The winning move is not build-everything or buy-everything; it is knowing which is which, and never giving away portability. Build where you have a real moat. Buy where you do not, and keep the model swappable so you keep control. The M.A.P. Test below gives you a ten-minute way to decide before you commit a budget to the wrong side.
Apple ran this calculation with a thousand engineers and a year of pain. You can run it on a whiteboard in ten minutes. Before your team commits budget to building an AI capability, put the decision through three gates.
M. Moat
Is this AI capability your actual product and differentiator, or is it plumbing? If customers choose you because of this exact capability, that is a moat, and a moat is worth building. If it is infrastructure every competitor also needs, it is plumbing, and you buy plumbing. Apple's moat is the device, the privacy stack, and the integration. The raw language model was plumbing, so Apple bought it.
A. Ability
Can you realistically match frontier quality with your data, your talent, and your budget, on your timeline? Be honest. If a frontier lab will out-ship you every quarter, building is a treadmill you lose. Redirect that spend to the layer you can win. Apple has more resources than almost anyone and still answered no. Your "no" is probably louder.
P. Portability
If you buy, what is the lock-in? Design for swap-ability. Put an abstraction layer between your product and the vendor, avoid single-vendor dependencies, and own the user relationship so you can change models without rebuilding everything. Apple kept portability by routing Gemini through Private Cloud Compute and opening iOS to multiple assistants. Buy the capability, keep the steering wheel.
Your Monday Morning Action
Pick the one AI capability your team is currently arguing about building. Run it through the three gates: Is it a moat or plumbing? Can you realistically win on quality? If you buy, how do you stay portable? Write one sentence per gate. If two of three say buy, stop building and redirect that budget this quarter. The companies that get this wrong spend a year building the thing a rival ships in a weekend.
Regulation
Both Flagship AI Laws Got Softer in Three Weeks. The Deadline You Dreaded Just Moved.
Colorado scrapped its own AI Act weeks before it was due to start: Gov. Polis signed SB 189 on May 14, repealing the original law and replacing it with a lighter, disclosure-based version effective January 1, 2027. In the EU, a Digital Omnibus deal pushes the heaviest high-risk obligations to December 2027. The August 2, 2026 EU date still stands, with penalties up to 7% of global turnover. The lesson: build to a portable standard like the NIST AI framework or ISO 42001, because the rules keep moving and whiplash is the only constant.
Adoption
ChatGPT Crosses 1 Billion Monthly Users, the Fastest Any App Has Ever Done It
ChatGPT hit 1 billion monthly active users in May, about three years after launch, outpacing the early growth of TikTok, Instagram, and YouTube, per Sensor Tower data reported by Reuters. AI assistants are no longer an early-adopter habit; they are a default tool for a billion people. If your product or team still treats AI as optional, your customers already disagree.
Infrastructure
Google Will Pay SpaceX $920 Million a Month to Rent AI Compute
Google agreed to pay SpaceX roughly $920 million a month for 32 months of compute capacity at xAI's data centers, with a similar Anthropic deal signed in May. Even as SpaceX's own Grok stumbles in government bids, the company found a business renting out the hardware. The capacity itself is now the asset, no matter whose model runs on it.
Big Tech
Apple Opens iOS to Claude and Gemini as Pick-Your-Own Assistants
Beyond the Gemini-powered Siri, iOS 27 Extensions let users set Claude or Gemini as their assistant, with an AI agent tier reportedly coming to the App Store. The phone is becoming a neutral platform for whichever model wins, not a walled garden for one. For builders, the distribution layer just cracked open.
~$1B/yr
Apple's Reported Annual Payment to Google for the Siri Model
Per Bloomberg's Mark Gurman. The price of buying the AI brain instead of building it.
1.2 trillion
Parameters in the Custom Gemini Model Apple Licensed Rather Than Built
The frontier-scale capability Apple decided was not worth building in-house.
1 billion
ChatGPT Monthly Active Users
Reached in May, the fastest any app has hit the mark, per Sensor Tower via Reuters.
Jan 1, 2027
New Effective Date After Colorado Repealed Its Original AI Act
SB 189, signed May 14, replaced the law weeks before its old June 30 start.
Google Gemini (Gemini API & Vertex AI)
The buy option that just won Apple
What it is: Google's frontier model family, available to any team through the Gemini API and through Vertex AI on Google Cloud. It is the same lineage Apple just licensed to run Siri, offered to developers and enterprises as a managed service with tuning, grounding, and governance built in.
Why it matters now: This is the clearest example of the "buy" side of the M.A.P. Test. When the company most famous for building everything in-house decides to call an API instead, that is the market telling you where the build-vs-buy line now sits. Gemini is the model enterprises, and now Apple, are integrating rather than rebuilding.
Who gets it: Available to any developer via the Gemini API and to organizations through Vertex AI. If you are weighing whether to train your own model, start here first: prove you cannot win with a bought model before you spend a year building one. Keep an abstraction layer in front of it so you stay portable.
"We didn't want to send users off into some chat experience in order to get things done."
Craig Federighi, Apple SVP of Software Engineering, describing Siri's design philosophy at WWDC 2025 · one year before Apple rebuilt Siri as exactly that, a chatbot, on Google's Gemini
Share AI News Insider
Know someone stuck in a build-vs-buy fight at work? Forward this issue.
SHARE & EARN REWARDS →AI News Insider · Your weekly edge in artificial intelligence
aiinsider247.com