When 'diversified' means two companies

1. Key Themes

---

Theme 1: Extreme Capital Concentration in AI — Just Two Companies Are Eating the Market

AI VC funding hit a record $407 billion in H1 2026 alone, surpassing all of 2025 ($264B). But the distribution is alarming in its concentration.

"OpenAI and Anthropic alone accounted for more than half of the H1 tally, totaling about $217 billion."

This means "diversified" AI exposure in a portfolio could effectively mean a two-company bet. The structural divide is stark: horizontal platform/frontier model companies captured 70.8% of deal value ($288B), while vertical application companies — which represent 62.9% of deal count — captured only 12.9% of capital.

---

Theme 2: AI Valuation Premiums Are Real in VC — But Unproven Beyond It

Late-stage AI companies are commanding extraordinary multiples relative to peers.

"AI companies raising Series D rounds and later are fundraising at 6.6x their non-AI peers, but whether these valuation premiums hold beyond VC is an open question."

The IPO test looms large. OpenAI is reportedly weighing a delay to 2027 "amid media reports that the company isn't meeting internal benchmarks and concerns about global market volatility." If public markets reprice AI at a discount to VC, the downstream implications for the entire stack are significant.

---

Theme 3: Macro Tightening Is Suppressing Deal Count, Not Enthusiasm

The decline in AI deal volume is being misread by many as a sentiment shift.

"The decline in deal count has more to do with the global macroeconomic backdrop than with any cooling of AI enthusiasm... The US Fed's scaling back of its rate-cut plans, in turn, has tightened credit and made investors more selective in their bets."

H1 2026 saw only 3,500 AI deals, down from 8,290 in all of 2025 — a dramatic drop. The implication: fewer, larger bets are being made, not an exit from AI investing.

---

Theme 4: Global VC Ecosystem Reshuffling — Beijing Rising, London Falling

City-level rankings matter for where founders build and where investors deploy.

"San Francisco continues to reign as the top venture ecosystem, with New York holding steady in second. Boston jumped into third place, pushing Los Angeles down to fourth, as Beijing climbed to fifth, knocking London down to eighth."

London's slide in the rankings is consistent with a broader PE/VC exit drought — "only eight UK-based companies backed by buyout firms going public over the past five years."

---

Theme 5: AI Infrastructure Buildout Carries Hidden Liabilities

Wall Street's largest money managers are betting the AI infrastructure market is not overbuilt, but there's a financial disclosure issue lurking.

"Hyperscalers look financially spotless, with little debt and top-tier credit ratings — but that's because $1 trillion in future data center payments simply isn't on the books yet."

This off-balance-sheet risk is a material consideration for investors evaluating hyperscaler health and the companies dependent on their continued capex spend.

---

2. Contrarian Perspectives

---

Perspective 1: The "AI Boom" Is Actually a Two-Company Phenomenon — Not a Broad Market

The surface narrative is that AI is the hottest investment category ever. The underlying reality is more concentrated and fragile than it appears. OpenAI and Anthropic alone absorbed $217B of $407B raised in H1 2026. Vertical application startups — the companies most operators and mid-market investors can actually access — got just 12.9% of capital despite representing 62.9% of deals. The "boom" in aggregate numbers masks a barbell market where the middle is being starved.

"The concentration of capital into frontier models shows the conviction investors have in OpenAI's and Anthropic's ability to not just scale their own platforms but the vertical applications built on top of them." — Dimitri Zabelin, PitchBook Senior Analyst

---

Perspective 2: Fewer AI Deals Is a Signal of Quality, Not Declining Conviction

A naive read of deal count falling from 8,290 (full year 2025) to 3,500 (H1 2026 only) would suggest the AI bubble is deflating. The article pushes back on this, attributing the decline to macro credit tightening, not waning interest.

"The decline in deal count has more to do with the global macroeconomic backdrop than with any cooling of AI enthusiasm."

For investors, this could signal a better vintage — fewer deals competing for capital means higher-quality companies are getting funded on more disciplined terms.

---

Perspective 3: Hyperscaler Balance Sheets Are Not As Clean As They Look

Conventional wisdom treats the major cloud/AI infrastructure players as financially fortress-like. The article flags that this perception is based on incomplete accounting.

"Hyperscalers look financially spotless, with little debt and top-tier credit ratings — but that's because $1 trillion in future data center payments simply isn't on the books yet."

This has direct implications for investors who assume hyperscaler strength creates a permanent floor under AI infrastructure demand.

---

3. Companies Identified

---

4. People Identified

---

5. Operating Insights

---

Insight 1: Vertical AI Startups Are Getting Deals But Not Dollars — Adjust Your Fundraising Strategy Accordingly

If you're building a vertical AI application, you're in the majority of deal activity but the minority of capital allocation. Vertical companies drove 62.9% of H1 deal count but only captured 12.9% of funding. This means operators in this space need to be more capital-efficient, demonstrate faster paths to revenue, and make a stronger case for why their platform won't be subsumed by OpenAI or Anthropic's own expanding vertical offerings.

---

Insight 2: Location Strategy for Startups and Investors Should Reflect Shifting Ecosystem Rankings

PitchBook's VC Ecosystem Rankings — built on six years of deal, exit, and fundraising data — now show Boston overtaking LA and Beijing overtaking London.

"The rankings... compare global cities based on the size and maturity of their startup networks, helping founders, operators and investors assess locations when deciding where to expand or invest."

For operators considering expansion or investors allocating to geography-specific funds, Boston's rise and London's fall to eighth are actionable signals worth weighting.

---

Insight 3: Macro Credit Conditions Are the Primary Lever on Deal Volume — Not AI Sentiment

"The US Fed's scaling back of its rate-cut plans, in turn, has tightened credit and made investors more selective in their bets."

Operators seeking to raise should recognize that the scarcity of deals reflects financing conditions, not lack of interest. Those who can demonstrate capital efficiency and near-term revenue will have a structural advantage in this environment.

---

6. Overlooked Insights

---

Insight 1: The US Defense Department Is Now a Major Balance-Sheet Backer of Deeptech

Sila, a silicon-carbon battery materials startup, received a $1.4 billion loan from the US Defense Department — a number that rivals many large VC rounds. This is a largely underreported channel of non-dilutive capital that deeptech hardware founders in energy, materials, and defense-adjacent sectors may be systematically underutilizing.

---

Insight 2: The Movie Theater Industry Is Quietly Recovering — and the Formula Is Specific

Tucked into the Side Letters section is a notable data point: five films have already crossed $1 billion in ticket sales in 2026, driven by "early 2000s nostalgia, happy endings and microbudget films." This is a signal with implications not just for entertainment investors, but for any operator thinking about what consumer emotional needs are driving spending behavior right now — a trend worth tracking across adjacent media and experience categories.