Monday, July 20, 2026
There's a lot of noise today about who's really winning the AI race, plus Apple and OpenAI are trading legal blows, and New York just hit pause on new data centers. Let's get into it.
Cheap Chinese AI models are catching up to the best from OpenAI and Anthropic
Two Chinese companies just made a lot of people nervous. Moonshot AI released a new model called Kimi K3, and Alibaba released an updated version of its Qwen model. Both companies say their models perform about as well as the top American models, like those from OpenAI and Anthropic, but cost a lot less to build and run. These models are also open source, which means anyone can download the underlying code and run it themselves instead of paying a subscription fee to a company like OpenAI. For years, the assumption in Silicon Valley was that the US had a clear lead in building the smartest AI. This week's releases suggest that lead is shrinking fast, and that gap in cost may matter just as much as raw smarts.
What this means for you: If a Chinese company can offer AI nearly as good for a fraction of the price, expect the tools you already use to get cheaper too, as competition heats up.
What this means for your business: Do not assume your current AI vendor is the only serious option. It is worth checking if a cheaper alternative now meets your needs, especially for high volume, lower stakes tasks.
Source: The Verge
Radar 01
Google is reportedly building its own chip to make its AI cheaper to run
Google's parent company Alphabet is working on a new computer chip built specifically to run its Gemini AI models more efficiently, according to a new report. Right now, most AI companies rely heavily on expensive chips made by Nvidia. If Google succeeds in building its own chip, it could run its AI for less money per task, which matters a lot as AI use scales up across the world.
What this means for you: The AI tools you use every day, including Google search and Gemini, could get faster and cheaper over time as this kind of behind-the-scenes cost cutting happens.
What this means for your business: Watch which cloud providers are investing in their own chips. Companies that control their own hardware may offer more stable pricing down the road.
Source: TechCrunch
Radar 02
Apple's lawsuit against OpenAI could complicate OpenAI's hardware ambitions
Apple has filed a lawsuit against OpenAI, and people are now debating what it means for OpenAI's plans to build its own hardware devices and eventually go public. OpenAI has been talking publicly about wanting to make physical gadgets, not just software. A drawn out legal fight with Apple, one of the biggest hardware companies in the world, could slow those plans down or scare off investors ahead of a potential stock listing.
What this means for you: If you were hoping for new OpenAI gadgets soon, this legal fight might push those launch timelines back.
What this means for your business: Legal disputes between major tech players can create openings for smaller competitors. Keep an eye on who moves into any gap this fight creates.
Source: TechCrunch
Radar 03
New York becomes the first state to freeze new AI data center construction
New York Governor Kathy Hochul has signed a one year pause on building new AI data centers in the state. Data centers are the huge warehouses full of computers that power AI. They use enormous amounts of electricity and water, and local communities have been pushing back. New York is the first state to actually stop new construction, and other states are watching closely to see if they should do the same.
What this means for you: If you live near a proposed data center site, this kind of policy could directly affect whether it gets built and how your local power grid and water supply are used.
What this means for your business: If your growth plans depend on more AI computing power, do not assume unlimited data center capacity will keep expanding. Regulatory pauses like this could slow supply and raise costs.
Source: Ars Technica
Try This Today
Before renewing or expanding any AI subscription, spend 20 minutes checking if a cheaper open source model, like the ones coming out of China, could handle your lower stakes tasks just as well. You might not need the most expensive option for everything.
Quick Hits
- Databricks, the data and AI company, just hit a $188 billion valuation, showing investors still believe big money is in helping companies actually put AI to work, not just build the underlying models. [1]
- Thinking Machines Lab, a new AI startup, released its first model called Inkling. It is a large, freely available model trained to understand video and audio, aiming to compete with OpenAI and Anthropic. [2]
- OpenAI's CFO introduced a simple scorecard for measuring whether AI spending actually pays off, focused on cost per completed task and how dependable the AI is, rather than just how impressive it sounds. [3]
Enjoyed today's read? Help me get this in front of more people like you.
Start every morning a little sharper.
Free. Takes 5 minutes to read. No spam.
