In five months, the valuation jumped from $740 million to $2 billion.
On April 29, Parallel Web Systems, the company founded by former Twitter CEO Parag Agrawal, announced a $100 million Series B round. Sequoia Capital led the financing, with Andrew Reed joining the board. Kleiner Perkins, Index, Khosla, First Round, Spark and Terrain, nearly all of the Series A backers, also returned.
Total funding has now reached $230 million, only five months after the Series A last November.
What Agrawal is building
This is not a search engine. It is not a Perplexity-style answer engine either.
Parallel is building a web index and API layer for AI agents. If an agent needs company filings, it calls one API. If it needs supply-chain data, it calls one API. If it is doing diligence research, it calls one API. Underneath is Parallel's own crawled web index, structured specifically for agent workloads.
Agrawal frames the bet this way:
We started Parallel from the belief that agents will use the web at a frequency one thousand times higher than humans, and that most of this work will happen in the background.
Translated into business terms: a person may open a browser 20 times a day; an agent could make 20,000 API calls. If that scale arrives, the old Google search model of paid clicks, SEO and ad auctions breaks down. Agents do not click ads. They need structured, low-latency data with verifiable provenance.
Whoever controls the web access point for agents controls a major part of the next search cash flow.
The customer list matters
| Company | How it uses Parallel |
|---|---|
| Harvey | Web data retrieval for legal reasoning use cases |
| Notion | Research performed by agents for users |
| Profound | Content research |
| Actively | Sales lead monitoring |
| Opendoor | Real-estate data queries |
| Several investment banks and hedge funds | Real-time market and company information |
| Two large U.S. P&C insurers | Claims-related verification |
More than 100,000 developers are already using the product. The common thread is clear: these are agent products meant to complete multi-step tasks autonomously, not simple chatbots.
Harvey's Gabe Pereyra put it plainly:
Parallel is how we bring the web into the picture.
Notion's Sarah Sachs added:
Like people, agents need to research the public web to do work and make decisions. Parallel makes that possible.
Why Sequoia doubled down
Andrew Reed is joining the board with a straightforward thesis:
Long-running agents are redefining products in every industry. Agents need the web.
Sequoia has already backed agent standouts such as Cursor and Harvey. Adding Parallel is a clear extension of that strategy: after betting on the application layer, it is now betting on the infrastructure layer.
It echoes Sequoia's cloud-native playbook from the 2010s, when it backed both Snowflake and Datadog. One side captures vertical applications; the other captures horizontal tooling.
Why not Bing or Google
Bing has an API. Google has Custom Search API. Brave, Exa and Tavily are also in the market. This is not an empty lane.
Parallel's pitch is that agent-first design changes the product requirements:
- Structured schemas: the output is not a lump of HTML, but JSON an agent can consume directly.
- Full provenance: each data point can be traced back to the original web page, a requirement for regulated sectors such as finance and law.
- Agent-grade latency: a human search engine can tolerate 500 ms. An agent chaining a dozen API calls in one task needs latency closer to below 100 ms.
The last point determines whether Parallel can serve production agents. If one task needs 50 web lookups and each takes 500 ms, the experience collapses into a 25-second wait.
What the bet is really on
Parallel also highlighted a detail that should make publishers pay attention: it is investing in open web economics and market mechanisms so content owners and data providers can earn from AI systems.
In plain terms, it wants to create a new paid pathway for data crawling.
That is the core fight between AI companies and media owners over the past year. Anthropic paid $1.5 billion to settle music copyright claims. OpenAI faces lawsuits from publishers. Reddit and Twitter have closed free API access. If Parallel can build a system that gives publishers a reason to opt in, it is not merely another search engine. It starts to look like Stripe for the agent era: a payment layer for data flow.
Is a $2 billion valuation expensive? That depends on whether this payment layer works. If it does, the story can be much bigger than a 10x markup.
Sources: CocoLoop, Parallel Web Systems hits $2B valuation five months after its last big raise (TechCrunch), Parallel Raises at $2 Billion Valuation to Scale Web Infrastructure for Agents (PR Newswire), Parag Agrawal's startup raises $100M to build a parallel web for AI agents (SiliconANGLE), Sequoia Capital leads Parallel's $100M raise at $2B valuation (TechFundingNews)