PLUS: OpenAI's new ad plans, Apple's AI pressure, and the industry's growing power problem
Good morning
A new all-optical chip is promising a major leap forward in AI computation, boosting speeds for complex vision tasks by over 100 times compared to current hardware.
This breakthrough uses light instead of electricity to compute, potentially solving the immense energy and time costs of running large visual models. Could this development unlock a new wave of real-time visual AI that is currently impossible with existing hardware?
In today’s Next in AI:
A 100x faster optical AI chip
AI's workaround for the power grid
OpenAI’s plan for ads in ChatGPT
Apple’s high-stakes AI pressure
AI's Light Speed Moment
Next in AI: Researchers have developed an all-optical chip named LightGen that performs complex AI vision tasks. The new chip boosts speed and energy efficiency by more than 100 times compared to top electronic chips.
Decoded:
LightGen integrates millions of photonic neurons on a single chip, using light instead of electricity to perform computations.
The chip excels at demanding generative vision tasks, including high-resolution image creation, style transfer, and even 3D generation.
This technology directly addresses the industry's growing computing power shortage, which has become a major bottleneck for developing larger AI models.
Why It Matters: This leap in optical computing could dramatically lower the immense energy costs and time required to run large visual models. It also paves the way for powerful, real-time visual AI applications that are currently constrained by hardware limitations.
AI's Unseen Power Play
Next in AI: AI's massive energy appetite is overwhelming America's power grid, forcing major labs like xAI and OpenAI to bypass it entirely and build their own onsite gas-powered generation to avoid multi-year delays and capitalize on a looming power crunch.
Decoded:
The grid is gridlocked, with utilities unable to approve new connections fast enough to meet demand. In Texas alone, tens of gigawatts of new data center load requests are submitted each month, while only a fraction gets approved.
Elon Musk's xAI pioneered the workaround with its 100,000-GPU cluster, which entirely bypassed the grid by using mobile gas turbines. This new strategy allowed the facility to come online in months instead of years.
The trend is creating a boom for unexpected industrial players. Thanks to data center demand, Caterpillar's power business has become its fastest-growing sales unit, prompting major factory expansions.
Why It Matters: This move to self-generation marks a fundamental shift in how AI infrastructure is built, prioritizing deployment speed above all else. It creates a massive new market for industrial power suppliers and signals a long-term challenge for the tech industry's energy supply chain.
Ads Are Coming to ChatGPT
Next in AI: OpenAI is reportedly planning to integrate ads directly into ChatGPT, a move that could give advertisers preferential treatment in conversations and fundamentally alter the user experience.
Decoded:
One internal discussion involved giving sponsored results priority, which could push more relevant, organic answers further down in the chatbot's response.
To avoid disrupting the user experience, OpenAI is exploring ideas like showing ads only after a second prompt, preventing an immediate bombardment of sponsored content.
With 900 million weekly users, the model could mirror today's sponsored search results, where users must often sift through ads to find organic information.
Why It Matters: This signals a major shift in how AI companies will pursue monetization, moving beyond subscriptions toward a massive advertising model. The core challenge for OpenAI will be balancing revenue goals without damaging the trust users have built with its tool.
Apple's AI Do-Over
Next in AI: After delaying its AI-Siri update in 2025, Apple faces a high-stakes 2026 launch. The company needs to deliver a breakthrough experience to drive major iPhone upgrades and prove it hasn't fallen behind its rivals.
Decoded:
Apple's strategy hinges on driving hardware sales, as its most advanced AI features require an iPhone 15 Pro or newer to operate.
A second miss could solidify the perception that Apple is woefully behind its peers in AI, risking its control over the next major computing platform.
Unlike rivals with subscription models, Apple's primary path to monetizing AI is through selling new iPhones, Macs, and iPads.
Why It Matters: This is a rare public stumble for Apple on a transformative technology. A successful launch in 2026 will not only drive sales but could also silence critics and reaffirm the power of its integrated hardware-software ecosystem.
AI Pulse
Instacart halted its controversial AI-driven price testing program, which showed different prices for the same items to different users, following public criticism and regulatory scrutiny.
The LLVM project proposed a new "human in the loop" policy for AI-assisted contributions, requiring that contributors fully review and understand any LLM-generated code before submitting it for review.
Multiple states enacted new laws for 2026 targeting the use of AI, with California and Oregon regulating AI in healthcare and Montana and South Dakota requiring disclosures for deepfakes in elections.
Pi Network launched Pi App Studio, a generative AI-powered coding tool to help non-technical creators build and deploy apps on its platform.