🚀 The Startup Khan Newsletter
September 24, 2026 — Bessemer Raises $5.75B as AI Rewrites Venture and SaaS Economics
🧠 Top Story: Bessemer Raises $5.75B to Invest Across the AI Stack
Venture capital is going full-stack.
Bessemer Venture Partners has raised $5.75 billion in new capital in a single close:
$1.75B → Seed and early stage
$4B → Growth stage
The firm says it has already invested roughly $3 billion across more than 260 AI-native companies since 2022, spanning compute, infrastructure, foundation models, developer tools, applications and agents.
The important signal isn’t simply the size of the fund.
It’s the structure.
Bessemer wants capital available when a company is discovering product-market fit—and when that same company needs hundreds of millions to scale.
Why founders should care
AI is changing venture economics.
The traditional path looked roughly like:
Seed → Venture → Growth investors → Public markets
But breakout AI companies can consume enormous amounts of capital while remaining private.
So large venture firms increasingly want to finance more of the journey:
Seed → PMF → Growth → Mega-rounds → Exit
That makes your earliest investor decisions more strategic.
A strong early investor may provide more than a $2 million check.
They can potentially provide:
Recruiting → Customers → Follow-on capital → Growth financing → Long-term institutional support
👉 Founder takeaway: The venture firm itself is becoming a full-stack financial platform.
Choose investors partly on who can still help when your company is 100x larger.
🤖 Funding Watch: Ema Raises $77M to Challenge the SaaS Seat
What happens when software stops selling access and starts selling completed work?
Ema has raised a $77 million Series B led by Creaegis, with Accel, Section 32 and Prosus increasing their investments.
Total funding now stands at $140 million.
Ema deploys teams of AI agents across:
HR → IT → Finance
to execute multi-step workflows across existing enterprise applications.
But its pricing model may be even more interesting than the technology.
Instead of primarily charging for:
Seats
or
Tokens
Ema says it charges around completed work and business outcomes.
The company reports more than 50 active enterprise deals, over 1 million active enterprise users, more than 5 million actions and queries handled, $150 million-plus in bookings and roughly 180% net-dollar retention.
Those are company-reported metrics.
And importantly, the $150 million figure includes multiyear contracted bookings.
It should not be read as ARR.
Why this matters
Traditional SaaS has a beautiful economic engine:
More employees → More seats → More software revenue
Agents challenge that assumption.
If AI can execute the workflow itself, the pricing model can shift toward:
Price × Work completed
or eventually:
Price × Business outcome
That expands the competitive market.
Agents aren’t only competing with software.
They’re competing with the labor used to operate the software.
👉 That’s potentially a much larger prize.
☁️ Infrastructure Watch: Modal Reportedly Eyes a ~$15B Valuation
Models are getting cheaper.
Making them useful in production isn’t.
AI infrastructure startup Modal Labs is reportedly discussing financing at approximately a $15 billion valuation.
Modal previously raised $355 million in May at a $4.65 billion valuation.
Its platform provides developers with compute and sandbox environments where code—including increasingly AI-generated code—can be safely executed and tested.
One important qualification:
This is not a completed financing.
The discussions have not been confirmed by Modal, and terms could change.
Founder signal
If financing eventually happens anywhere near the reported valuation, it would reinforce where investors still see scarcity.
Not necessarily in access to another model.
But in:
Compute → Execution → Sandboxing → Orchestration → Deployment
AI infrastructure is moving above the GPU.
👉 The chip provides compute. The software around it makes that compute usable.
🧰 Tool of the Day: Modal
Today’s infrastructure story is also useful at the founder level.
Modal gives developers serverless compute infrastructure designed for demanding workloads, including AI inference, batch processing and GPU-intensive applications.
For AI startups, that’s useful when you need significant compute without immediately building an infrastructure team around it.
Potential uses include:
- Model inference
- GPU workloads
- Batch processing
- AI-generated code execution
- Sandboxed environments
- Scheduled jobs
👉 Best use: When your startup needs serious compute but managing serious infrastructure isn’t your differentiation.
Infrastructure abstraction lets small teams operate above their weight.
📊 Trend Check: SaaS Pricing Is Moving Toward the Outcome
Watch how software companies charge.
There’s an important progression happening:
SAAS
Sell access.
↓
COPILOT
Sell productivity.
↓
AGENT
Sell completed work.
↓
OUTCOME
Sell the business result.
Moving down that stack can dramatically increase the market a startup attacks.
But it also increases responsibility.
A SaaS company can charge because the customer logged in.
An outcome-based company gets paid because something actually happened.
That changes product design.
It changes pricing.
It changes margins.
And eventually, it could change how enterprise software companies are valued.
🌟 Founder Spotlight: Follow the Budget, Not the Category
Suppose you’re building an AI agent for accounts receivable.
If you describe it as software, you’re competing for the company’s software budget.
But if your agent actually performs collections work, you’re potentially competing with:
Software spending + Outsourcing + Internal labor
That’s a much larger economic pool.
The same logic applies across:
Customer support
Accounting
Legal
Recruiting
Sales operations
IT
👉 Founder question: What budget disappears when your product succeeds?
That’s often a better way to understand your true market than asking which SaaS category you belong to.
🔴 Founder Opportunity Radar
🚨 TODAY — Tech Policy Press Fellowship
Applications for the 2027 Tech Policy Press Fellowship close September 24.
The year-long, part-time program targets mid-career journalists, researchers and professionals working on technology-policy questions.
⏰ TOMORROW — TechCrunch Disrupt Pricing Deadline
Current TechCrunch Disrupt 2026 ticket pricing changes after September 25 at 11:59 p.m. PT.
The conference runs October 13–15.
📣 September 30 — Disrupt Exhibitor Deadline
Startup exhibition bookings remain open through September 30 at 11:59 p.m. PT.
For B2B founders actively fundraising or selling, exhibiting may be more useful than simply attending.
🌱 NOW OPEN — VILLAGEx 2027
The Idea Village has opened applications for its 2027 VILLAGEx accelerator, targeting Gulf South startups preparing to raise capital.
⚡ Quick Win: Run the Budget Replacement Test
Pick your product’s most important customer workflow.
Then ask:
What does the customer spend today to get this outcome?
Include everything:
Software
Employees
Contractors
Outsourcing
Management overhead
Errors and delays
Now compare that total with your pricing.
You may discover you’re competing for a much larger budget than your software category suggests.
🧩 Final Thought
Today’s strongest signals connect.
Bessemer is raising billions to finance AI companies across their lifecycle.
Ema is experimenting with selling completed work instead of software access.
Modal represents the infrastructure required to make increasingly autonomous software actually run.
AI isn’t merely changing software features.
It’s changing:
How startups are funded.
How software is priced.
Which workflows technology owns.
Which budgets technology competes for.
That’s a much deeper transition than adding AI to SaaS.
The next great enterprise company may not sell the software seat.
It may sell the work that seat used to enable.
— The Startup Khan
Your 5-minute startup edge — funding news, tools, and founder playbooks.
