From the 24 AI companies with over 100 million in financing, the three main lines of the AI application explosion are clearly seen.

In an emerging industry, investment often speaks volumes. After all, nothing is more convincing than "putting real money on the line."

According to TechCrunch statistics, as of June 18th, 24 AI startups in the United States have completed single - round financing exceeding $100 million. On the surface, this figure is almost the same as last year.

However, if we break down the details, we'll find that the capital flow this year is revealing three important signals:

Firstly, "super - financing" is contracting. There were 7 financings at the $1 billion level last year, but this year, only OpenAI has crossed this threshold, indicating the squeezing effect of leading players.

Secondly, the AI application layer is bidding farewell to the "diverse development" stage and entering a phase with a "clear main line." Programming, healthcare, and law are becoming the directions where capital is most concentrated, and the industry's implementation path is becoming clearer.

Thirdly, the AI investment paradigm is shifting from model mania to "ARR victory." Although the overall investment in the current AI field is still far higher than the return, a group of companies have stood out. Under the bonus of "cheap computing power + integrated workflow," they have polished their products and are running out truly sustainable business models. (ARR stands for Annual Recurring Revenue)

Capital doesn't lie. The directions they bet on are often the directions the industry is heading.

01 Programming/Healthcare/Law Become the Golden Scenarios for AI Implementation

In 2024, a total of 49 AI startups in the United States completed single - round financing exceeding $100 million, including 7 "super deals" with single - round financing exceeding $1 billion. As of June 18th, 2025, 24 companies have completed financing at the $100 - million level, with a total financing amount exceeding $49 billion.

Judging from the capital flow, the basic layer remains strong, and computing power and models are still "hard currency."

Model companies such as OpenAI, Anthropic, and Together AI (received $305 million in Series B) continue to attract a lot of capital; infrastructure - related companies have raised a total of $1.03 billion, accounting for 25% of the financing transactions in the first half of the year, making them the top players among the "shovel sellers":

TensorWave (AI chips, $100 million in Series A)

CelestialAI (optical interconnection technology, $250 million in Series C)

Lambda (GPU cloud services, $480 million in Series D)

EnChargeAI (edge AI chips, $100 million in Series B)

SnorkelAI (data annotation, $100 million in Series D)

LMArena (large - model benchmark testing, $100 million in seed round)

Compared with the "stability" of the basic layer, the application layer is being rapidly reconstructed. In the past year, the AI application layer has undergone a transformation from "diverse development" to a "clear main line." Three tracks are becoming the "golden scenarios for AI implementation": programming, healthcare, and law.

AI programming is one of the earliest scenarios to achieve PMF (PMF stands for Product - Market Fit). So far, two AI programming companies have raised more than $100 million, namely:

Anysphere (the parent company of Cursor, raised $9 billion)

Turing (AI training and programming, $111 million in Series E, valued at $2.2 billion)

Interestingly, among the five AI programming companies that raised over $100 million last year (Poolside, Magic, Cognition, Codeium, Augment), except for Codeium, there has been no news of new - round financing for the rest.

The frequent change of leading players also reflects that this track is still in the early stage of rapid evolution and has not yet stabilized.

AI in law and healthcare have become the high - value directions most favored by capital in vertical scenarios, accounting for 20% of the total number of related financings.

Taking AI in law as an example, Harvey and Eudia completed large - scale financings of $300 million and $105 million respectively this year.

Harvey's legal AI assistant, Harvey Assistant, can automatically perform core tasks such as legal research, document drafting, contract analysis, and due diligence. Built on a large - language model, it can integrate court documents, provide legal insights with citations, and assist in proofreading drafts, translation, and filling in contract fields, and is widely used in law firms and corporate legal processes.

Eudia focuses on building an AI agency system for corporate legal teams, specializing in contract review and compliance automation, and has been implemented in several Fortune 500 companies such as Cargill.

Together, they represent a trend shift in the legal industry: AI is moving from a "legal tool" to a "quasi - legal employee," not only saving time but also directly integrating into professional processes.

Healthcare is also one of the most imaginative vertical scenarios for AI implementation. This year, three AI healthcare companies have completed financing exceeding $100 million:

Abridge (medical transcription, $250 million in Series D): It focuses on automatically transcribing doctor - patient conversations into structured clinical records and has been implemented in more than 100 medical systems across the United States.

HippocraticAI (medical large - model, $141 million in Series B): It builds a medical large - model for clinical scenarios, supporting tasks such as patient triage and decision - making assistance.

Insilico Medicine (AI - driven drug discovery, $110 million in Series E): It uses AI to drive new drug discovery and molecular design and has end - to - end drug development capabilities.

They represent the ways AI cuts into the three key links of medical information collection, clinical decision - making, and drug R & D, and also reflect the industry's transition from "pilot exploration" to "systemic reshaping."

This wave of AI healthcare fever has also been verified at the capital level.

Bessemer pointed out in a research report: "More than 70% of institutions in the global healthcare industry are actively exploring the implementation of AI, and three - quarters have increased their IT budgets in the past year; among all the healthcare technology startups that received financing in 2024, nearly 40% have deployed AI as their core capability."

02 The Shift of AI Application Investment Paradigm, from Model Mania to "ARR Victory"

OpenAI's post - investment valuation reached $300 billion in March 2025, nearly doubling compared to October 2024; Anthropic also rose from $40 billion to $61.5 billion in 5 months.

However, behind the investment of billions of dollars, the profitability and commercialization paths are still being questioned.

Meanwhile, a group of AI application companies are quietly achieving "double - high - speed growth" in valuation and revenue.

Taking Harvey in AI law as an example, its valuation soared from $3 billion to $5 billion in just 4 months;

Cursor completed two rounds of financing in half a year, and its valuation directly skyrocketed from $2.5 billion to $9.9 billion.

Lovable, a newly - emerged company, also reached a valuation of $1.5 billion in half a year.

The key lies in that AI applications are no longer just conceptual stories. Under the combination of low - cost computing power + high - value workflow, they are quickly building real and sustainable business models.

As David Cahn, a partner at Sequoia, said: A group of AI - native application companies have polished their products under the bonus of "cheap computing power + integrated workflow." Companies like Harvey and Sierra are building truly sustainable business models.

Specifically, among the US AI application companies that have raised more than $100 million, several have achieved amazing ARR growth:

Anysphere: Its revenue increased by 230% in 3 months, reaching $500 million

Codeium: Its revenue tripled in 2 months

ElevenLabs: Its revenue increased by 370% in 10 months

If we focus on the core scenarios, the explosion of AI applications is even more remarkable.

According to the "Investment Internship Institute," 4 AI programming companies currently have an ARR exceeding $100 million, and 3 others are between $50 million and $100 million.

Among the AI programming products with an ARR exceeding $100 million, in addition to Windsurf, which was acquired by OpenAI, there are also Cursor, GitHub Copilot under Microsoft, and Augment, incubated by Sutter Hill Ventures.

Some investors estimate that the overall ARR of AI programming products currently reaches about $1.5 - 2 billion and is expected to rise to $3 - 4 billion by the end of the year.

From the financing rhythm to the revenue growth rate, AI applications are telling a different story from large - models. They don't rely on burning money to expand but on real user needs and high - frequency and essential scenarios to quickly polish products, accumulate capabilities, and release commercial value.

When the "computing - power bonus" gradually stabilizes, the real competition in AI will return to product and implementation capabilities. And AI applications may be the first group to break through the bubble and move towards certainty.

03 Top Venture Capital Firms Bet Collectively, Lightspeed Venture Partners Is the Most Aggressive

As AI applications move beyond the tool stage towards system integration and industry reconstruction, a group of leading venture capital firms are increasing their bets, and capital is quickly flowing into high - value vertical scenarios such as enterprise search, AI law, and AI healthcare.

In the financing cases of these 24 US companies, Lightspeed Venture Partners is the most active investor, participating in the investment of 6 companies, including Glean (enterprise search), Anthropic (large - model), Abridge (healthcare), Snorkel AI (AI data platform), Nexthop AI, and Reflection AI.

a16z is one of the most active institutions this year, having invested in 4 star companies, including Anysphere (AI programming), LMArena (model evaluation), ElevenLabs (audio generation), and HippocraticAI (medical large - model), covering the key paths from model evaluation to industry implementation.

Sequoia Capital has been a bit more restrained this year, only betting on Harvey (AI law) and Glean (search), targeting the certain tracks that are "implementable and monetizable."

Meanwhile, industrial capital has also accelerated its entry into the market.

Nvidia not only promotes the ecosystem at the chip end but also strengthens the "AI upper - layer application binding" of its computing - power advantage by investing in emerging companies such as SandboxAQ, Together AI, and Lambda; Salesforce Ventures has bet on Anthropic and Together AI, trying to integrate large - model technology into its CRM workflow system.

It can be seen that the investment logic in the AI application field has changed significantly: from blindly "betting on technology" to precisely "investing in scenarios"; from burning money to grab users to valuing commercialization efficiency.

The collective bets of top VCs and industrial capital are a vote of confidence in the "AI implementation ability." This wave of AI application companies can not only develop products but also run out a commercial closed - loop, becoming the truly penetrating innovation engines in the next stage.

This article is from the WeChat official account "Crow Intelligence Talk," written by Smart Crow, and is published by 36Kr with authorization.