The startup ledger is large enough to look healthy and concentrated enough to look brittle. Crunchbase’s digest-linked H1 analysis says global venture investment reached $510 billion in the first half of 2026, surpassing the full-year 2025 total. The digest also says US and Canadian startups accounted for $392 billion of that amount.

The headline number is not the whole story. The digest says OpenAI and Anthropic alone represented $217 billion of the global total, or 43% of all venture funding. That is not a normal distribution of risk capital. It is a capital market organizing itself around a small number of companies viewed as infrastructure for the next computing era.

For founders, this has two readings. The generous version is that a large platform shift creates room for supporting software, services, data, security, compliance, workflow redesign, and domain-specific products. When frontier platforms raise enormous sums, entire ecosystems can form around their capabilities and gaps.

The harsher version is that capital abundance at the top does not automatically make fundraising easier below. Investors may become more selective about anything that looks like a wrapper, a feature, or a product likely to be absorbed by a model provider. A non-frontier AI startup has to prove why it owns distribution, workflow depth, data advantage, regulatory trust, or a painful niche that a platform will not prioritize.

Wonder’s reported $650 million Series D at a $9 billion valuation shows capital still reaching outside pure model infrastructure. Food technology, delivery operations, and kitchen logistics remain hard physical systems where software can improve utilization but cannot erase execution risk. Walden Robotics, emerging with $300 million according to the digest, sits closer to the automation frontier: physical AI aimed at industrial and logistics deployment.

The reported SpaceX-Cursor deal is the file to treat most cautiously unless confirmed by primary parties and closing documents. The digest frames it as a record all-stock acquisition expected to close in Q3. If true, it would mark a striking convergence of aerospace, developer tooling, and AI-assisted code production. Until then, it is a high-signal claim that still needs a hard verification stamp.