The EU orders Google to open Android and share its search data with rivals

Information Technology 5 min 80 sources

Brussels invokes its gatekeeper law to force open two of tech's biggest platforms, while TSMC's record quarter, a Coca-Cola ransomware shutdown, and China's chip and model push fill out a busy day.

Key takeaways

The EU ordered Google to open Android and share search data with rivals by 2027 — a rule change aimed at the data advantage that keeps it dominant, not a one-off fine.
TSMC posted record profit and pledged another $100 billion for US factories even as chip stocks slid — the companies building AI's physical capacity and the market pricing it are answering different questions.
A ransomware attack halted Coca-Cola's Fairlife dairy production, a reminder that a breach can stop a physical factory line, not just leak data.

The biggest story in tech today came out of Brussels, not Silicon Valley. On Thursday the European Commission ordered Google to give rival search engines and AI assistants “comparable access” to Android and to some of Google Search’s data [5][9][11]. Google has until January 2027 to start sharing search data and until July 2027 to change how Android works [5].

Brussels reaches for the machinery, not the fine

The two decisions come out of the EU’s Digital Markets Act, or DMA — a 2024 law that names a handful of dominant platforms as “gatekeepers” and requires them to give competitors the same access to their systems and data that they give themselves [5][11]. The point of the law is structural, not punitive: instead of a one-off fine, it changes the rules a company has to operate under going forward [11].

What Google actually has to do is the interesting part. Rival AI assistants and search engines must be allowed to plug into Android and reach parts of Search that were previously walled off [5]. That reaches directly into Google’s AI plans — its Gemini assistant loses some of the home-field advantage of living on the operating system most of the world’s phones run [5]. Google warns the changes could weaken user privacy and security, the standard industry response to forced data-sharing [11].

The timing sits inside a wider European anxiety. A New York Times piece the same day laid out Europe’s push for “technological sovereignty” — the wish to run critical software, chips, and AI on infrastructure the continent controls, and why that is proving so hard when the dominant platforms are all American [18]. Google, for its part, was handed a grace period to comply; Apple, in a separate DMA ruling, must make its Siri AI compliant before launch, not after [6].

The angle: if you build products or integrations for the European market, the interoperability requirements arriving in 2027 change what you can assume about who owns the default search box and the assistant slot on a phone. That is a dependency worth mapping now.

The chip engine keeps building, even as the stock keeps wobbling

TSMC, the Taiwanese company that manufactures most of the world’s advanced chips, reported second-quarter profit up 77% and said it will pour an additional $100 billion into its Arizona factories [3]. The company was explicit about the driver: demand for the chips that train and run AI models [7]. This is the picks-and-shovels layer of the AI boom — whoever wins the model race, the silicon still gets fabricated in the same few plants.

That confidence ran straight into a nervous market. US chip and memory stocks slid again, the S&P 500 and Nasdaq fell as chip names weakened, and in Asia SoftBank dropped 8% tracking the Wall Street AI sell-off [0][2][22]. The split is worth holding in your head: the companies building the physical capacity are reporting record orders while the share prices swing on doubt about whether the AI demand behind those orders will last. A record quarter and a falling stock are answering two different questions — one about today’s bookings, one about tomorrow’s expectations.

When ransomware stops a factory line

Coca-Cola disclosed to US regulators that its Fairlife dairy subsidiary was hit by ransomware — malicious software that scrambles a company’s files and demands payment to unlock them — and that it had suspended production “for the foreseeable future” [20]. This is the part of a cyberattack that gets less attention than stolen data: the physical line stops. Milk isn’t bottled because the systems that run the plant are locked.

There was a win on the other side of the ledger. UK police said the jailing of two young hackers — Owen Flowers, 18, and Thalha Jubair, 20, who pleaded guilty to breaching Transport for London in 2024 — had “severely” hampered Scattered Spider, one of the most active cybercrime crews of recent years [21]. And a cautionary tale from The Register: a law firm ran on a single master password that let anyone log in as any staff member or client and read their personal and health records — a hole the IT worker who found it was told not to touch [12].

The angle: the Fairlife shutdown is the argument for treating operational technology — the systems that run a physical process — as seriously as the servers holding customer data. A breach that halts a production line is a business-continuity event, not just a privacy one.

The model race widens, and Google stumbles

China’s Moonshot AI unveiled what it called the world’s largest open AI model — one whose weights anyone can download and run on their own machines — and said it was closing the gap with US labs [14]. Its next model, Kimi 3, is expected to perform at or near the level of Anthropic’s Opus 4.8, according to reporting cited by TechCrunch and the Financial Times [19]. Open models matter because they let companies run capable AI on their own hardware instead of renting it through an American lab’s service.

Google, meanwhile, delayed the launch of a new Gemini model after it fell short of internal targets, according to Bloomberg [32][41]. Two data points, one week: the incumbent slips a release while a Chinese challenger ships an open model. The frontier is no longer a single-file lead.

The under-covered story: China builds the memory it can’t buy

Far from the headlines, retail investors in Shanghai bid for shares in CXMT — China’s main maker of memory chips — at more than 200 times the number on offer [10]. The frenzy is a signal, not just a number. Memory is one category where China still leans heavily on foreign suppliers, and export controls have made that dependence a strategic weakness. A wildly oversubscribed listing is the market pricing in a national push to build a homegrown memory industry — the same sovereignty instinct driving Europe’s story today, playing out in a different capital.

02 · Lesson · why it matters

Why regulators are draining Google's flywheel instead of breaking it apart

A dominant company's real lead is often a loop that feeds itself — and every time you use the thing, you give the loop another push.

The strange shape of the remedy

Look closely at what the EU actually told Google to do this week. It did not order the company split in two. It did not cap its market share or ban a product. It told Google to share some of its search data with rivals, and to let other search engines and AI assistants plug into Android.

That is an odd-looking punishment. If Google is too powerful, why hand its competitors a copy of its homework instead of just making Google smaller? The answer is that Brussels is not aiming at Google’s size. It is aiming at the machine underneath the size — the thing that made Google big in the first place and would make it big again even if you cut it in half tomorrow.

The advantage that builds itself

That machine is a loop. Engineers call it a flywheel: something that spins faster the more force you feed it, and then keeps that speed on its own.

Google’s loop runs like this. More people search, so Google sees more of what people click and ignore. That data lets it rank results better. Better results pull in more people. More people generate more data. Round and round. Each turn of the wheel makes the next turn easier.

The important thing about a flywheel is that the advantage is not a fixed object you can point to and remove. It is not one brilliant algorithm sitting in a vault. It is a process — a self-reinforcing loop where being ahead is what keeps you ahead. The lead is not the cause of the dominance. The lead and the dominance are the same wheel, seen at two moments.

What looks like “just better” is often “just earlier”

Stand in front of Google and it feels simple: it gives good answers, so people use it. That story is true. The results really are good, and they genuinely help the person searching. Hold on to that — the loop serves the user, not only its owner.

But it is not the whole shape. A newcomer with an equally clever team cannot catch up, because it starts with no data, so its results are worse, so it draws fewer users, so it gathers less data. The gap is not a gap in talent. It is a gap in history. Google is not only better; it is earlier, and earliness compounds. What presents itself as plain quality — “they just make the best product” — sits on top of a structure that was set before today’s contest began, and that structure quietly decides who is allowed to compete at all.

You cannot cut a loop in half

Now the remedy makes sense. Break Google into two companies and you get two smaller flywheels, each still spinning, each still pulling data-rich results out of its own turning. Within a few years one of them wins and the loop is whole again. Splitting the product does nothing to the process.

So regulators reach for the two things that actually slow a flywheel. First, bleed off what spins it — force some of the search data out to rivals, so a newcomer’s wheel is not starting from a dead stop. Second, lower the walls around it — make Android let other assistants in, so being the default is no longer an automatic push. Neither move attacks Google’s quality. Both attack the loop that turns quality into a permanent lead. Whether it works is an open question; Google has until 2027 and says the data-sharing risks users’ privacy. But you can finally see what the target is.

The wheel is made of us

Here is the part that is easy to miss. A flywheel this size is not spun by Google. It is spun by everyone who searches. Your queries, your clicks, the results you scrolled past — all of it is force fed into the wheel that keeps the incumbent ahead of the next challenger. You are not watching the loop from outside. You are one of the hands on it, and so is nearly everyone you know.

This is not a reason to feel guilty for using a search engine. It is a description of how these advantages actually form: not by one grand act, but by billions of small, free, sensible choices, each of which feels like nothing. That is why the loop is so hard to see from any single seat. From where you sit, you just looked something up. The wheel it turns is only visible from far above, in a regulator’s filing or a rival’s frustration — and even they can only see part of it. The most powerful structures in the systems we live inside are often the ones assembled, one ordinary click at a time, by the very people they hold in place.

03 · Lab · your turn

Slow the Flywheel

Rehearse why regulators drain a data loop and open a platform instead of breaking a company apart to slow a self-reinforcing advantage.

04 · Hope · carry this

No lead is permanent once people learn to see the loop that built it. The slow, unglamorous work of naming how advantage really compounds — and reaching for the lever that actually loosens it — is how every open door in this industry got pried open in the first place.