Overview of AI investment in the first half of 2025: 58% of global venture capital flows to AI

Author: Catalaize

Compiled by: Felix, PANews

Global investment in AI startups from January to June 2025 far exceeded that in the first half of 2024. The first quarter of 2025 alone attracted about $60 billion to $73 billion, more than half of the total for the whole of 2024, and a year-on-year increase of more than 100%. In the first quarter, venture capital investment in AI companies accounted for about 58% of the total, compared with about 28% a year ago. This clearly shows the "AI FOMO" mentality of investors.

This means that capital is concentrating on the AI field on an unprecedented scale, and major institutions will double down on companies that are believed to be able to win in the AI field, which may reshape the capital allocation pattern in the second half of the year.

Huge financing led by a few giants

This period was particularly marked by very large late-stage rounds, led by leading companies. In March, OpenAI raised $40 billion (the largest private round ever) at a valuation of $300 billion, while Anthropic’s $3.5 billion Series E round valued it at $61.5 billion. Several other deals, such as Safe Superintelligence’s $2 billion round and Neuralink’s $650 million Series E round, further increased the total.

What this means is that a winner-takes-all dynamic is concentrating most of the capital in the hands of a very small number of companies, squeezing out funds that might otherwise flow to earlier-stage or smaller companies.

Transaction size under the barbell effect

In addition to those high-profile mega-rounds, mid-sized deals surged, while seed round activity remained selective. The median seed round in the AI field reached about $15 million (average about $41 million), and the median Series A round was about $75 million to $80 million, both of which are much higher than the historical average (the median Series A round in all industries worldwide in 2022 was about $10 million). The median growth-stage financing of Series C and Series D is concentrated between $250 million and $300 million, while the average is pulled up by extreme cases such as OpenAI.

What this means: The ballooning deal sizes reflect fierce competition for industry leaders. Investors who can’t write nine-figure checks may turn to niche or earlier-stage investments, while any startup that claims an AI narrative can get larger rounds and higher valuations.

Industry and geographic concentration

Players in the generative AI and core models/infrastructure sectors attracted over $45 billion in funding in the first half of the year, accounting for more than 95% of total disclosed funding. Applied AI verticals were relatively underfunded (~$700 million in medical/biotech; ~$2-3 billion in fintech/enterprise). Geographically, the United States (especially Silicon Valley) dominates: over 99% of global AI funding in the first half of the year went to companies headquartered in the United States. Asia and Europe lagged behind, with China’s largest deal (Zhipu AI) raising $247 million; while Europe saw only a few mid-sized rounds (e.g., UK-based Latent Labs raising $50 million).

What this means: The boom is centered in the United States and led by a few large companies; governments and investors outside the United States are expected to respond in the second half of the year to avoid being left behind by setting up national AI funds, providing incentives, or making cross-border investments.

Outlook for the second half of the year: enthusiasm is high but caution remains

Despite record capital investment, investor prudence is returning. Many financing rounds in the first half of the year focused on strategic or corporate investors (cloud service providers, chip manufacturers, defense companies), indicating that investors prefer projects with practical application scenarios and strategic synergies. Entering the second half of the year, investors will pay close attention to the performance of startups that have received huge funds in product delivery, revenue, and response to regulations, especially in the context of increasingly fierce competition.

What this means: Capital in the second half of the year is likely to favor companies that demonstrate efficiency and real market appeal - especially "tool and shovel" suppliers (tools, chips, enterprise software), which will raise the barrier to entry for new entrants, consolidate the advantages of existing companies, and pose a challenge to new entrants.

importance

The first half of 2025 is a make-or-break moment for AI investing. The current flood of money into AI—and its skew toward a few players and regions—will shape the innovation landscape and competition for years to come. For investors, understanding where the money is flowing and why will be critical to navigating the second half of 2025. Will the winners justify their valuations, or will there be a pullback and refocus? H1 data provides early clues to portfolio strategy, policy considerations (such as antitrust and national security issues), and founders’ fundraising prospects over the next six months.

The most noteworthy financing in the AI field in the past month

Macro and trend analysis

1. Financing momentum: unprecedented surge year-on-year

In the first half of 2025, venture capital investment in AI startups far exceeded the same period in 2024. Reliable data shows that about $70 billion flowed into AI companies in the first quarter alone, more than half of the total funding in the AI field in 2024. This means that the funding amount in the first half of 2025 is more than twice that of the first half of 2024 (in US dollars).

In the first quarter of 2025, AI's share of global venture capital jumped to about 53% to 58%, compared with about 25% to 30% a year ago. This means that more than half of the world's venture capital is currently invested in the field of AI.

Driver: A handful of mega-rounds; without them, global venture funding was roughly flat year-over-year.

Impact on H2 2025: Overall venture capital metrics are likely to depend on deal flow in the AI space; any cooling of enthusiasm in the AI space could pull down overall funding levels.

2. Financing stage: Late-stage financing has increased significantly, while early-stage financing is mixed

Data shows that the transaction scale in the AI field is distributed in a barbell shape.

Late-stage (C+ rounds) dominate: Late-stage financing across all industries totaled $81 billion in the first quarter of 2025, up approximately 147% year-over-year, with AI being the main driving force.

• The average size of Series D and E rounds is approximately $300 million to $950 million (with a median of approximately $250 million to $450 million).

Early Stage: The number of deals declined (globally early stage deals fell by about 19% year-over-year), but the size of financing increased significantly.

• The median seed round for AI startups in the first half of 2025 was about $15 million; Lila’s $200 million seed round was an outlier.
• The median Series A round is approximately $75 million to $80 million.

The takeaway: Investors are putting money into fewer, bigger bets—feeling confident in specific AI themes and cautious about others. This polarization should continue in the second half of the year.

3. Industry configuration: basic model and infrastructure construction

About 95% of AI funding is chasing generative AI model developers and their infrastructure (cloud computing, chips, development platforms). OpenAI and Anthropic alone attracted about 60% of AI funding in the first half of the year.

In comparison, the vertical application areas are trivial:

• Healthcare/Biotech AI: ~$700M (e.g., Hippocratic AI raised $141M, Insilico raised $110M).
• Financial services and business productivity: just a few billion dollars combined.
• Robotics/Defense AI: Niche but worth watching (e.g., Shield AI raised $240M).

Investor logic: Control the "AI stack"; vertical applications may become commoditized (note: the unique value of the brand and other original values of the commodity disappears due to full market competition) or face a longer GTM cycle.

4. Geographic distribution: concentrated in the United States, with the Bay Area accounting for half of the financing amount

In the first quarter, 71% to 73% of global venture capital went to North America; by value, about 99% of funding in the AI field is concentrated in the U.S. The San Francisco Bay Area alone (including OpenAI) accounts for nearly half of global venture capital.

EMEA: Only a few mid-sized AI deals (Latent Labs raised $50 million, Speedata raised $44 million).

Asia Pacific: Only $1.8 billion was raised for AI in Q1 2025 (down 50% year-over-year); the largest round in China was $247 million raised by Zhipu AI.

Bottom line: The United States has an advantage in funding in this “AI arms race.”

5. Investor landscape:

Sovereign wealth funds and cross-border funds (Saudi Arabia's Prosperity7, Malaysia's Khazanah, Thrive Capital) led multiple rounds of financing.

The corporate venture arms of large tech companies (Microsoft, Salesforce, Google) are very active.

Net effect: capital pours in from all sides.

Forward-looking outlook for the second half of the year:

Regulatory milestones

Governments are still figuring out how to deal with AI. In the EU, the AI Act is expected to be finalized by the end of 2025. In the second half of the year, startups are expected to launch lobbying battles and early compliance signals may appear. In the United States, executive orders on AI and any movement in Congress—hearings, proposed legislation—will be crucial. New regulations around data use, model transparency, or chip export controls could reshape the economics of startups and investor confidence.

• Positive Expectations: Clearer, business-friendly guidelines legitimize the use of AI in various industries.
• Negative expectations: Tough rules (e.g., liability for AI mistakes) could scare off startups and investors.

Also, keep an eye on U.S. government AI procurement—rumors of a multi-billion dollar program could provide important demand signals for enterprise-focused AI companies.

IPO channels and exit routes

Despite a surge in private equity funding in 2025, we have yet to see a breakout AI IPO. This may change in the second half of the year. Companies like Databricks, Stripe (AI-related) and even OpenAI could be potential IPO candidates.

• A successful IPO could potentially repricing the market, unlocking liquidity at later stages and providing comparable data.
• The continued IPO slump could shake investors’ confidence in AI startups’ exit timelines.

Meanwhile, M&A activity is likely to escalate. Big tech companies could make moves: Google, Microsoft, or Nvidia could acquire smaller AI teams or core infrastructure vendors. A major AI acquisition could reshape the competitive landscape and generate returns for venture capital firms.

Technology Breakthroughs and Product Launches

Expect big news to be revealed: perhaps the next generation of models from OpenAI, or hardware from a collaboration between Sam Altman and Jony Ive.

Any major breakthrough in capability (e.g., a model that can reason or a model that costs 10x less) could justify high valuations and trigger a new wave of capital.

Also keep an eye on enterprise traction – API sales, SaaS adoption and revenue. But there is also the risk that a security incident or public misuse could lead to a regulatory backlash that could dampen sentiment.

In short, technical and commercial execution in the second half of the year will determine whether the optimistic trend in the first half of the year can be sustained.

Regulatory and ethical resistance

If governments or the public feel AI is out of control, expect swift intervention: perhaps through licensing, fines under the General Data Protection Regulation (GDPR), or strict restrictions on certain models.

Ethical headwinds: Scandals, mass layoffs due to automation, or AI-generated misinformation could quickly change market sentiment, making it harder to commit money.

Computing and Talent Limitations

The lifeblood of AI — graphics processing units (GPUs) and elite engineers — remain scarce.

GPU bottlenecks can force underfunded teams to quit, while well-funded companies hoard computing resources.

The war for talent is intensifying, with OpenAI and Google scrambling to recruit top talent.

Burn rates are skyrocketing: Some startups are spending more than $100 million per year on cloud services, but are failing to launch products quickly enough. If the gap between costs and products continues to widen, expect financing discounts and brutal market resets.

Model Commercialization

Ironically, the Large Language Model (LLM) competition is driving rapid commoditization. Open source releases (Meta's LLaMA, Mistral, etc.) blur the distinctions.

Moats are shifting toward data quality, distribution channels, or vertical integration.

If OpenAI starts losing ground to a slimmed-down pool of open source players, or to models developed in-house, venture capitalists might rethink what “defensibility” actually means.

The second half of the year may serve as a wake-up call: not every finely tuned wrapper deserves a $1 billion valuation.

Forecast for the second half of 2025

Financing scale slowed down but remained high

After the first half of the year, the pace of dealmaking will slow. No more $40 billion rounds are expected, but quarterly AI funding will still be double 2024 levels. The boom is still going on, just more robust.

Major liquidity event is coming

Expect at least one $10 billion+ exit: an IPO (e.g. Databricks) or an acquisition by a legacy company trying to maintain relevance.

This will impact investor sentiment and reset pricing expectations.

The startup ecosystem is clearly divided into different parts. By the fourth quarter, the differentiation will be obvious:

The top 5-10 AI companies (with strong funding and momentum) will gradually exit and may recruit talent through acquisitions.

Those middling or overhyped startups that haven’t yet achieved product-market fit? Many will pivot, experience valuation cuts, or simply fade away.

Investors will reward execution that generates revenue, not just research proposals or GPUs.

Final Conclusion

The next six months will stress-test the AI narrative. Will 2025 be the beginning of lasting change, or a bubble that needs to be corrected?

Some bubbles will burst, but the core argument remains valid. AI remains the most attractive frontier for venture capital, but the flow of funds will be more cautious.

Related reading: Building the cornerstone of the AI economy: How does AI reshape the stablecoin landscape?