DX Today AI Daily Brief - Wednesday, August 12, 2026

DX Today AI Daily Brief - Wednesday, August 12, 2026

Intel closes an upsized $20 billion share sale to bankroll its chip manufacturing buildout, while Tencent posts 11 percent revenue growth on AI powered advertising even as heavy AI spending squeezes profit. CoreWeave more than doubles revenue and reports a $104 billion backlog with capacity effectively sold out, and CME Group moves to make computing power a tradable commodity with the first regulated compute futures launching in October. River AI, founded by xAI co founder Igor Babuschkin, lands $1.1 billion to build an open AI stack, Mistral pledges a gigawatt of European compute by 2030, and IBM inks a $240 million inference deal with Together AI. Ryanair signs a five year Google Cloud partnership to bring AI tools to 35,000 employees, Microsoft cuts the price of its upgraded MAI Code coding model by roughly 75 percent, SpaceXAI launches Grok Bot agents that work like always on digital coworkers, LTX ships a faster than real time open weights video model, and Manus returns to independence after Chinese regulators force Meta to unwind its $2 billion acquisition.

DX Today - Your daily AI news briefing, delivered every morning.

#AINews #ArtificialIntelligence #AIInfrastructure #AIAgents #DXToday

Transcript

SPEAKER_04 0:00

It's Wednesday, August 12th, 2026. You're listening to the DX Today AI Daily Brief. Today, Intel closes an upsized $20 billion share sale to fund its chip build-out. Computing power becomes a tradable commodity as CME Group announces the first regulated compute futures. And AI startup Manus goes independent after Beijing forces Meta to unwind its $2 billion acquisition. Let's get into it.

SPEAKER_00 0:32

Intel has closed the largest equity raise in its history, pricing an upsize $20 billion common stock offering at $95 per share. The chipmaker initially set out Monday to raise $15 billion, then expanded the deal Tuesday after investor demand reportedly topped $100 billion. It is Intel's first common share sale in more than five decades, covering roughly 210 million new shares. Proceeds are earmarked for Intel's manufacturing build-out, including its 14A process node, as the company positions its foundry business to serve surging AI chip demand. Shares initially fell almost 5% on dilution concerns before the scale of institutional appetite became clear, a sign of just how much capital markets are willing to fund the AI hardware race.

SPEAKER_02 1:21

Now to earnings in Asia. General and administrative expenses jumped 22% on research and development for its foundation models and AI features across its apps. The results capture the tension running through big tech earnings this season. AI is clearly lifting revenue, particularly in targeted advertising, but the infrastructure and talent bills are arriving first, and investors are watching closely for when that spending turns into durable profit growth.

SPEAKER_03 2:22

Next, the AI Cloud. AI Cloud provider Corweave delivered second-quarter revenue of about $2.58 billion, up 112% from a year ago and ahead of Wall Street expectations. The adjusted loss came in at $1.14 per share, better than the consensus forecast, and adjusted operating income reached $128 million. The headline number was the backlog. CoreWeave now reports roughly $104 billion in future contracted revenue, and Chief Executive Michael Intrator told investors that near-term capacity remains effectively sold out. Shares jumped around 14% after the report. The company still carries heavy debt to fund its data center expansion, but the quarter reinforced that demand for AI computing capacity continues to outrun supply. From capacity to commodity markets.

SPEAKER_01 3:18

AI computing power is officially becoming a tradable commodity. CME Group, the world's largest derivatives marketplace, will launch the industry's first regulated compute futures contracts on October the 5th, pending regulatory approval in partnership with benchmark firm Silicon Data. The contracts will track the hourly rental cost of Nvidia's H100 and B200 graphics processors, letting AI developers, cloud providers, and investors hedge or lock in the price of computing capacity the way airlines hedge jet fuel. Silicon Data separately raised $30.5 million to expand the pricing benchmarks behind the market. If it works, compute joins oil, gold, and grain in the institutional trading framework. And the cost of intelligence gets its own forward curve.

SPEAKER_04 4:07

Now a billion-dollar bet.

SPEAKER_05 4:09

River AI, the startup founded by former ex-AI co-founder Igor Babushkin, has raised $1.1 billion just two months after emerging from stealth. General Catalyst and AMPPBC led the round, with participation from Nvidia, AMD Ventures, and Y Combinator. River is building what it calls an open AI stack, tools that let enterprises and developers train, fine-tune and fully own custom models on their own data rather than renting intelligence from a handful of frontier labs. Babushkin has said he wants AI that is trainable and not controlled by any single big company. The size of the round for a company this young shows investors are still writing enormous checks for credible open alternatives across the Atlantic to Europe.

SPEAKER_00 5:02

French AI company Mistral laid out a sweeping plan for European computing sovereignty. The company says it will build one gigawatt of AI compute capacity in Europe by 2030, offer service-level guarantees on regional inference endpoints so customer workloads stay on the continent, and sell five-year contracts denominated in what it calls European compute units, locking in enterprise customers now. Notably, Mistral will also host a leading Chinese open model on European infrastructure, a signal that sovereignty means independence from both American and Chinese clouds rather than isolation. Analysts note the capital requirements are immense, with next generation AI data centers costing tens of billions of dollars. But Mistral is betting European governments and enterprises will pay a premium for guaranteed local capacity.

SPEAKER_02 5:51

Next, an infrastructure alliance. IBM signed a $240 million multi-year agreement with Together AI, the cloud company specializing in open source models. Under the deal, IBM will deploy a dedicated cluster of Nvidia's latest HGXB 300 systems on IBM Cloud, connected with high-speed Spectrum X networking, to run large-scale inference for Together's customers. The partnership gives Together dedicated enterprise grade capacity, and it gives IBM a marquee AI-native customer for its cloud infrastructure business. Both companies pitched the deal as making open source model serving cheaper and more reliable for enterprises, a segment where demand has been climbing steadily as companies look for alternatives to closed frontier models and seek more control over their AI costs.

SPEAKER_03 6:52

From clouds to airlines. Europe's largest airline is going all in on Google. Ryanair signed a five-year data and AI partnership with Google Cloud that will roll out Google workspace and cloud services to 35,000 employees across its network. The airline plans to use Gemini Enterprise to support decision making, optimize crew logistics across its flight operations, and build custom AI agents for functions across the business. Ryanair frames the technology overhaul as core to its plan to carry 300 million passengers a year by 2034. For Google, the deal is another flagship enterprise win in a European market where cloud providers are competing fiercely to attach AI tools to large legacy operations, from banks to carriers to national governments.

SPEAKER_04 7:42

Onto developer tools.

SPEAKER_01 7:44

Microsoft released MAI code 1.1 Flash, an upgraded version of its in-house coding model, and is rolling it out inside GitHub Copilot. The new model adds native vision support, so it can understand images like screenshots and diagrams, streams tokens about 25% faster, and uses about 25% fewer tokens to complete the same task. Those efficiency gains let Microsoft offer it at roughly one quarter the price of its predecessor. A cut of about 75%. Support for the original MAI Code 1 Flash ends September 10th. The aggressive pricing underlines how quickly the cost of AI coding assistance is collapsing as Microsoft pushes its own models and competes with fast-improving open alternatives.

SPEAKER_04 8:34

Now agents get computers.

SPEAKER_05 8:36

SpaceXAI, the artificial intelligence division of SpaceX, formerly known as XAI, launched Grokbot in beta. The product turns Grok from a chat assistant into a team of always-on agents, each with its own persistent computer in the cloud that can sign into applications, browse the web, keep context over time, and keep working even when your laptop is closed. Users can hand off tasks like managing inboxes, research, and multi-step workflows to agents that operate software much the way a human coworker would. The beta is available on desktop and iOS for subscribers to the company's top Grok tier, along with premium tiers of the cursor coding platform, with enterprise access to follow.

SPEAKER_04 9:25

Next, faster video generation.

SPEAKER_00 9:27

Video AI company LTX released LTX 2.5, the newest version of its open weights video and world model. The headline capability is speed. The model can generate a 10-second AI video from a single image in under seven seconds on Nvidia's latest superchips, effectively faster than real time. It ships with native integration into Comfy UI, the node-based tool that has become the standard prototyping environment for visual AI workflows. Beyond video generation, the company positions the release as a world model, one that simulates how scenes evolve over time with applications in robotics training as well as media production. Because the weights are open, developers can download the model, customize it, and run it on their own hardware.

SPEAKER_02 10:14

And finally, a breakup. AI agent startup Manus says it will soon resume operations as an independent company, closing the book on one of the most closely watched deals in the industry. Meta announced in December that it would acquire the Chinese founded company for roughly $2 billion, but Chinese regulators stepped in this spring and demanded the deal be unwound. Manus said Tuesday the separation is now in its final stage. As part of the transition, data generated by some users since late December will be deleted to satisfy regulatory requirements in certain jurisdictions. The episode has become a cautionary tale about geopolitics sitting in the middle of AI deal making, with both Washington and Beijing increasingly willing to block cross border ownership of frontier AI companies.

SPEAKER_04 11:14

That's your briefing for Wednesday, August 12th, 2026. For DX today, stay curious.