Top 10 AI News July 18 2026: Google's Big Miss Google delayed Gemini again as its stock fell, Kimi K3 topped the coding charts, and Apple passed Nvidia. Today's top 10 AI stories, explained simply.

AI News Today July 18 2026: Top 10 Stories

The most hyped AI day of the year did not go how Google planned. The Gemini launch everyone waited for got delayed again, Google's stock dropped, and while it stumbled, a Chinese open model quietly topped the coding charts and China's president launched a global AI club with 29 countries. Oh, and Apple passed Nvidia to become the most valuable company on Earth. I read everything so you only need five minutes. Here are today's top 10 AI stories, in plain English.

1. Google's Big Gemini Launch Flopped and Its Stock Dropped

Google's Gemini 3.5 Pro, the model the entire industry expected to launch on July 17, got delayed again after it reportedly fell short on coding and reasoning in testing, and Alphabet's stock dropped about 4 percent on the news. To be precise, Google has still not published any official details, so the honest summary is this: as of July 17, there is no confirmed Gemini 3.5 Pro, just a rebuilt model that reportedly still is not good enough to ship.

This is a genuine surprise, because expectations were enormous. Google had already scrapped the first version in June and restarted training, and the second attempt reportedly still trails Anthropic's Fable 5 and OpenAI's GPT-5.6 on exactly the coding and reasoning tasks that businesses pay for. A 4 percent drop in Alphabet's value is investors reacting to a hard truth: being late is one thing, but being late and still behind is what turns a delay into a real question about whether Google can keep up at the frontier.

Here is the fair other side. Refusing to ship a flawed flagship is actually the responsible call, and Google still has the deepest research team in AI and Search reaching billions of people, so counting it out would be silly. But the timing could not be worse, landing the same day a Chinese model topped the charts and China launched a global AI organization.

My take: shipping something broken to hit a date would have been worse. But by delaying, Google handed its biggest moment of the year to its rivals, and in AI, momentum is everything. This was a rough day for the company that used to set the pace.

2. A Chinese Open Model, Kimi K3, Won the Day Instead

While Google stumbled, Moonshot AI's Kimi K3, launched the night before, shot to number one on the Frontend Code Arena, a leaderboard that pits models head to head on real coding tasks, with a 76 percent win rate that beat Anthropic's Fable 5. It also scored 88.3 on a tough coding benchmark called Terminal-Bench. In hours, a Chinese open model did what Google could not do all week: post a top result on the record.

What makes this a big deal is the word open. Kimi K3 is a massive 2.8-trillion-parameter model, the largest open-weight release ever, and its weights will be free to download by July 27. That means anyone, from a startup to a student, will be able to run a model that just beat one of the best paid models in the world at coding. Its pricing to use through an API is also far below the top closed models. It is not perfect, ranking only ninth on general chat, so it is a coding and agent specialist rather than an all-rounder, but on the tasks it wins, it wins big.

The timing was clearly no accident. On the exact day Google wanted the spotlight, a Chinese lab grabbed the top coding spot with a model it is about to give away for free, hours before China's president took the world AI stage. That is coordination, and it worked because the model is genuinely good.

My take: Kimi K3 is the real winner of the week. When a free model beats a top paid one at coding, the whole question of why you should pay for AI gets a lot harder to answer. This is the story to watch.

3. China's President Launched a Global AI Club With 29 Countries

Chinese President Xi Jinping used his first-ever keynote at the World AI Conference in Shanghai on July 17 to launch the World Artificial Intelligence Cooperation Organization, or WAICO, a new international body headquartered in Shanghai, with 29 countries including Pakistan, Russia, and Kazakhstan signing up as founding members. Xi described AI as something that should be, in his words, a symphony of global cooperation rather than a solo performance by one country, and pushed for fair access for all nations.

This is bigger than a speech. A real organization with a headquarters and 29 member countries is actual diplomacy, and Xi paired it with a strong pitch for open-source AI and a promise to help developing countries get access, while criticizing the US for restricting technology. The message to countries locked out of American AI and chips is simple: build on China's tools, join China's club, and get the access China is offering to share. It is the rulebook version of Kimi K3 winning the same day.

The West has no equal answer ready. Europe is writing safety rules and the US is holding meetings, but neither has proposed a global club that other countries can actually join. In fact, Google's own AI boss called for a US-led group the same week, which quietly admits the West is behind on this.

My take: whoever writes the rules shapes the game for everyone. China just set up the table and invited the world to sit down, and right now nobody else has sent an invitation. That matters more than any single model.

4. Apple Became the Most Valuable Company in the World

Apple overtook Nvidia to become the world's most valuable company, approaching a $5 trillion value, with Nvidia dropping to second place. It caps a strong run for Apple, which just got approval to bring its AI to China using Alibaba's models and is building its own AI chips. For two years Nvidia, the company selling the chips behind the AI boom, was the undisputed king, so Apple retaking the crown is a real shift in the story.

Think about what this signals. Nvidia sells the shovels in the AI gold rush, and its rise to the top was treated as the defining chart of the era. Apple passing it suggests investors are starting to reward the companies that put AI in front of billions of real people over the one that supplies the raw hardware, or at least spreading their bets. Apple ships AI to more devices than anyone, makes money through the most profitable ecosystem in tech, and is now building its own chips to depend less on Nvidia.

The honest catch is that these market-value crowns change hands all the time, and Nvidia could take the top spot back on its next strong earnings report, especially since chip demand is still booming. But the deeper signal holds: getting AI to users, not just building the hardware, is looking like the more durable way to make money.

My take: the big question in AI is no longer who builds the best model, it is who makes the money once models become cheap and everywhere. Apple just gave a very loud answer to that question.

5. Anthropic Filed to Go Public, and Could Be Worth Over $1 Trillion

Anthropic, the company behind the Claude AI models, filed confidential paperwork to go public with an IPO possible by late 2026, and investor interest could value it at over $1 trillion. Anthropic has quietly become the revenue leader in AI, on track for roughly $47 billion a year and reportedly already profitable, driven mostly by its Claude coding tools and business customers.

A trillion-dollar value for a company that did not exist five years ago would be extraordinary, and it reflects a specific bet: that Anthropic's careful, business-first, safety-focused approach earns steadier money than flashier consumer rivals. The company topped this month's AI safety report card, keeps winning famous researchers, and has locked in the computing power to keep its costs predictable, which is exactly what stock-market investors love. The contrast with OpenAI, heading to its own IPO while fighting Apple's lawsuit, could not be sharper.

The caveat is that filing paperwork is the start of a long process, not a guaranteed price, and a trillion-dollar debut needs the markets to stay excited through a busy, crowded autumn of AI listings. But the direction is clear, and it reframes the whole race.

My take: Anthropic spent 2026 making boring, disciplined moves while everyone else made headlines. Boring and disciplined is exactly what wins over stock-market investors. If it lists near $1 trillion, the careful strategy wins.

6. Mira Murati Gave Away a Huge AI Model on a $2 Billion Budget

Thinking Machines, the startup founded by former OpenAI technology chief Mira Murati, released Inkling, a huge 975-billion-parameter AI model that anyone can download and use for free, and the company reportedly raised a $2 billion seed round, one of the largest ever. Murati helped build the most famous paid models in the world at OpenAI, and her first move on her own is to give a powerful one away.

The numbers make this serious, not symbolic. A 975-billion-parameter model is frontier-scale, and a $2 billion budget means Thinking Machines can afford to train and run it without charging for every answer right away. Releasing it openly, from someone with Murati's background, is a strong signal about where she thinks the value in AI is heading. It lands in the same stretch as Kimi K3 and DeepSeek, meaning several of July's most talked-about models are free to download, and now one comes from a star American founder.

The obvious question is how the company makes money if the model is free, and the likely answer is the same one European lab Mistral uses: sell services, custom training, and hosting around the free core. For anyone learning to build with AI, more powerful free models simply means more to experiment with at no cost.

My take: when the person who ran engineering at OpenAI raises $2 billion to give a model away, the debate about free versus paid AI is basically settled at the top. The paid labs now have to explain what you are paying extra for.

7. DeepSeek Is Now Worth $74 Billion and Slashing Prices 75 Percent

Chinese AI company DeepSeek is reportedly raising over $70 billion at a $74 billion value, up from around $50 billion before, and preparing to list on the Shanghai stock market next year. Its whole strategy is aggressive pricing, roughly 75 percent cheaper than rivals, which keeps pressuring the entire market. Its actual revenue is a modest $400 to $500 million a year, so the huge value is a bet on the future, not today's sales.

The jump in value shows how seriously the market now takes Chinese open models. DeepSeek has topped the free-model leaderboards for months, its next big version lands July 24, and its rock-bottom prices are the number the whole industry gets compared against. A $74 billion value on half a billion in sales is a bet that free, cheap models will dominate the huge volume of everyday AI work, and that DeepSeek will be the default choice for it. Listing in Shanghai keeps that value inside China, matching this week's pattern of the AI world splitting in two.

The honest tension is that cutting prices 75 percent is a weapon that also limits your own income. It works if massive scale pays off later, and it fails if prices never recover. It is the classic playbook of losing money now to own the market later.

My take: DeepSeek is spending its profits to win the standard, and between it, Kimi K3, and Murati's Inkling, the free-model team now has the models and the money to force every paid lab to defend its prices.

8. The Boss of DeepMind Says AGI Could Arrive in Five Years

Demis Hassabis, the Nobel Prize winner who runs Google DeepMind, said that artificial general intelligence, meaning AI that can match humans across most tasks, could arrive within five years. He also called for an international watchdog and a US-led group to check powerful AI models before they are released. Coming from one of the most respected and usually cautious leaders in AI, in the same week Google delayed its own model, the comment carries real weight.

The five-year timeline matters mostly because of who said it. Hassabis is known for being measured, not hype-prone, so him putting AGI within five years suggests the people closest to the research see things speeding up. His call for a watchdog is the more practical part, and it landed the same day China launched its WAICO club with 29 countries already signed up. The contrast is striking: China built an actual organization with members, while the West is still asking for one to exist.

The fair caveat is that AGI predictions have a long history of being wrong, and five years is a guess from someone with a stake in the field's momentum, not a certainty. But the governance point stands no matter the timeline. If the leading Western AI boss is publicly asking for a watchdog that does not exist yet, the fact that it does not exist is a real problem.

My take: the most important thing Hassabis said was not about AGI, it was admitting the West needs a group it does not have yet. China just showed everyone how far ahead it is on exactly that.

9. AI CEOs Are Getting Bodyguards After Threats

AI company executives are getting increased physical protection after a rise in threats, including an attempted firebombing at OpenAI CEO Sam Altman's home and assassination warnings against industry leaders. It is a grim milestone: the people building AI are now considered targets serious enough to need security details. Threats and violence are never acceptable, and this is a sobering sign of how high the emotions around AI have climbed.

The development reflects how quickly AI has gone from a tech-industry topic to a raw public nerve. As AI reshapes jobs, concentrates huge wealth in a few hands, and stokes fears from layoffs to bigger worries, the executives at the center have become lightning rods for anger. It connects to a pattern we have tracked all month: factory workers striking over robots, most workers wanting AI profits shared, and now physical threats against founders. The tension between the industry and the public it is changing keeps rising.

The responsible way to read this is not about any one person, but about what it reveals: the conversation about who benefits from AI and who pays the price has broken down badly enough to turn into threats. That is a warning the industry should not brush off.

My take: when the people building a technology need bodyguards, something has gone wrong in how that technology is being explained and shared. Fixing that conversation is now a safety issue, not just a PR one.

10. The Open-Model Wave Is Quietly Taking Over

Step back from the individual stories and July 17 looks like a real turning point. The West's most anticipated launch flopped, a Chinese open model topped the coding charts, China launched a global AI club, and several of the month's biggest model releases, Kimi K3, DeepSeek, Murati's Inkling, and the Bonsai model that runs on a phone, are all free to download. For at least a week, the momentum shifted away from expensive paid models and toward free ones and Chinese institutions.

The big picture is that AI is now a genuinely multi-sided contest, fought across models, chips, money, and rules all at once, with no single company or country controlling everything. The US still leads in the very best paid models, business revenue, and stock markets, with Anthropic's giant IPO and Apple's $5 trillion crown as proof. China leads this week in free-model momentum and governance. And the free-model camp, spanning both countries, is quietly winning the argument on price and access, which affects everyone who uses AI.

The honest read is that one bad week does not decide a decade, and Google is far too strong to write off. But July 17 punctured the idea that the frontier belongs permanently to a few big paid Western labs, and that idea may not fully recover. The thing to watch for the rest of July is the free-model wave, with Kimi K3 and DeepSeek weights arriving within days.

My take: if free models keep beating paid ones, the whole business of charging for AI has to be rewritten. That is the real story of late 2026, and it sped up a lot this week.

Frequently Asked Questions

Q: Did Gemini 3.5 Pro launch?

No. Reports indicate Google delayed Gemini 3.5 Pro again on July 17 after the rebuilt model fell short on coding and reasoning in testing, and Alphabet shares dropped about 4 percent. Google has published no official model details, so leaked specs like the 2-million-token memory remain unconfirmed.

Q: Is Kimi K3 better than ChatGPT for coding?

On one key coding leaderboard, yes. Moonshot AI's Kimi K3 reached number one on the Frontend Code Arena with a 76 percent win rate and scored 88.3 on Terminal-Bench, a top-tier coding result. It ranks lower on general chat, so it is a coding specialist, and its open weights are due by July 27, 2026.

Q: What is WAICO?

WAICO is the World Artificial Intelligence Cooperation Organization, an intergovernmental body headquartered in Shanghai that Xi Jinping announced at the World AI Conference on July 17, 2026. It launched with 29 founding countries including Pakistan, Russia, and Kazakhstan, positioning China as a leader in global AI governance.

Q: Is Apple the most valuable company in the world?

As of July 17, 2026, Apple overtook Nvidia to become the world's most valuable company, approaching a $5 trillion market value, with Nvidia in second. These rankings shift with daily trading, but it signals investors rewarding AI reaching real users, not just chip supply.

Q: Is Anthropic going public?

Anthropic filed confidential paperwork for a potential IPO by late 2026, with investor interest that could value it over $1 trillion. It is the AI revenue leader at roughly $47 billion a year and reportedly already profitable, driven by its Claude coding tools and enterprise customers.

Q: How much is DeepSeek worth?

DeepSeek is reportedly raising over $70 billion at a $74 billion valuation, up from around $50 billion, while preparing a Shanghai stock listing next year. Its revenue is an estimated $400 to $500 million a year, and its pricing runs about 75 percent below rivals.

Q: What is Thinking Machines' Inkling?

Inkling is a 975-billion-parameter open AI model released by Thinking Machines, the startup founded by former OpenAI technology chief Mira Murati, which reportedly raised a $2 billion seed round. Anyone can download and customize it, making it a major bet on free, open models.

Q: When will AGI arrive?

Google DeepMind CEO Demis Hassabis said on July 17 that artificial general intelligence could arrive within five years, and called for an international watchdog to vet powerful models. AGI timelines are uncertain and have often been wrong, so treat any specific date as a forecast, not a fact