NeuralStackThe biggest AI and tech rounds tracked in the latest funding roundup show a market still dominated by enormous checks, but with a clearer center of gravity: infrastructure, frontier research and the industrial systems underneath AI. The tracker says it has verified and sourced billions across dozens of rounds, and the largest names on the list are not app startups. Instead, they are companies building the compute, energy and model layers that power the sector.
At the top of the roundup is Prometheus, described as the AI venture co-founded by Jeff Bezos, with a $12 billion Series B co-led by JPMorgan Chase and BlackRock. DeepSeek follows with roughly $7 billion in its first external funding round, backed by a group that includes Tencent, CATL, the National AI Industry Investment Fund, NetEase and JD.com. Anthropic also appears among the biggest raises, with AMD committing up to $5 billion in a strategic equity investment tied to a supply agreement.
Safe Superintelligence is another headline deal, with Nvidia reportedly committing about $5 billion in an investment paired with a long-term compute partnership on Nvidiaโs Vera Rubin platform. The tracker frames that as the clearest example of a broader supplier-as-investor pattern, where chip and cloud sellers back the very companies that buy their hardware and services. In that view, some of the money is effectively cycling back into the infrastructure stack that the AI boom depends on.
Below those megadeals, the list continues to tilt toward the physical and operational layers of AI. Joulent raised $1.75 billion for AI datacenter energy and power infrastructure, while Baseten closed a $1.5 billion Series F for AI inference. Fireworks AI also drew a large round for inference and fine-tuning, and Together AI continued the push into training and serving infrastructure for open-source models.
Defense technology is another major destination for capital in the roundup. Helsing raised $1.8 billion at an $18 billion valuation, Shield AI closed a $1.5 billion Series G, and Quantum Systems raised $1.2 billion. The funding tracker treats these as part of the same mid-2026 pattern: large checks are flowing to companies that combine AI with autonomy, sensing and hardware.
Prediction markets and fintech-adjacent infrastructure also show up near the top, with Kalshi raising $1 billion at a $22 billion valuation and Taktile drawing attention for automated risk decisioning. Even there, the common thread is infrastructure rather than pure software distribution. The trackerโs own framing is that the most valuable AI businesses in this phase are the ones supplying the rails for other companies, not just building end-user features.
Why this matters
The funding mix matters because it shows where investors think the durable profits in AI will land. If the largest rounds continue to favor chips, power, inference, defense and model infrastructure, then the competitive advantage in the sector may accrue first to the layer that controls compute access and deployment economics.
It also suggests that AI fundraising is becoming more interconnected. Strategic money from Nvidia, AMD and other suppliers does not just signal confidence; it can lock in future demand for their own products and cloud capacity. That creates a feedback loop in which the biggest backers are also the biggest beneficiaries of the spending they help enable.
The trackerโs latest table suggests that this is not a one-off wave. The concentration in infrastructure-heavy rounds spans frontier labs, datacenter power, autonomous systems and inference platforms, and the ranking has room for more deals as daily coverage adds new raises. If the current pattern holds, the next phase of AI fundraising may look less like a broad startup boom and more like a capital race around the industryโs bottlenecks.
The round list is still evolving, but the signal is already clear enough: mid-2026 AI money is chasing the stack, not the slogan. That leaves the app layer under pressure to prove it can command the same scale of capital as the systems that make those apps possible.