This is today's edition of The Download, our weekday newsletter that provides a daily dose of what's happening in the world of technology.
These Startups Are Chasing the Next Big Thing in LLMs
Nine years after Google researchers introduced the transformer, this family of neural networks has become the engine inside nearly every major large language model. But transformers are starting to show their age. As LLMs grow bigger and more capable, transformers have emerged as a bottleneck. Their dense attention mechanism becomes increasingly expensive as the amount of text expands, and they struggle to keep track of large amounts of information simultaneously.
In response, a wave of startups is pursuing novel architectures that could solve these limitations. Here are four new ideas for how to address the transformer problem—innovations that could reshape LLMs for good, making them faster, far more efficient, and potentially even smarter.
—Will Douglas Heaven
This story is for subscribers only. If you're not subscribed, consider signing up to access the full article and more insights from our editorial team.
How AI Academic Research Is Shifting
In a separate development, the landscape of AI academic research is undergoing a significant transformation. Increasingly, leading researchers are moving from university labs to industry positions, drawn by greater resources and data access. This shift raises important questions about the sustainability of academic AI research and its impact on long-term innovation.
As of 2026, this trend has accelerated, with several prominent AI labs at tech companies outpacing university research in terms of published breakthroughs. However, academic institutions are fighting back by fostering partnerships and focusing on foundational research that industry may overlook.
—Additional reporting by The Download team
Nvidia Secures $500 Billion from Wall Street for AI Infrastructure
In other news, Nvidia has secured a massive $500 billion investment from Wall Street to expand its AI infrastructure. This funding aims to support the company's efforts to build more advanced data centers and computing systems needed to train the next generation of AI models. The investment underscores the growing financial commitment to AI infrastructure as demand for powerful processors and specialized hardware continues to surge.
The deal, announced amid a broader AI investment boom in 2026, positions Nvidia to further dominate the AI hardware market, but it also raises questions about the environmental and economic costs of rapidly scaling up AI capabilities.