Render's $1.5B Valuation Hides a $0.15/GB Question: What 100% Growth Doesn't Tell You About Your Next Bill

Render raised $100 million at a $1.5 billion valuation in February 2026 on more than 100% year-over-year growth. Two months later it cut the bandwidth that ships with a plan from 100 GB to 5 GB on Hobby and 25 GB on Pro, with every gigabyte above that now $0.15. Every legacy workspace auto-migrates by August 1. The valuation headline and the repricing footnote landed in the same quarter — and only one of them tells you what your app will cost next year.

If you read the funding news as "Render is healthy, my stack is safe," here is the number you actually came for. Take the canonical indie stack — one web service, one background worker, one managed Postgres — on Render Pro. The base is $95 a month before you serve a byte. At 500 GB of outbound traffic, that stack costs $166.25 on Render. The same containers on one Hetzner CX22 with 20 TB of included bandwidth cost €3.79. At 1 TB, Render is $241.25; Hetzner is still €3.79. The growth number says nothing about which of those two numbers you will pay — and the repricing is what decides.

The rest of this post shows where each line of both bills comes from, sweeps the traffic from 25 GB to 2 TB so you can find your own row, and explains why a valuation is the wrong signal for the question you are actually asking: is next year's growth coming from more tenants like you, or from the same tenants paying more per GB?

What Render actually raised​

On February 17, 2026, Render announced a $100 million extension to its Series C. The round was led by Georgian, which also led the original Series C, with participation from Addition, Bessemer Venture Partners, General Catalyst, and 01 Advisors. Post-money valuation: $1.5 billion. Total funding to date: $258 million.

The press release pitch was not "cheaper Heroku." It was "the cloud for AI-native software" — runtime for AI agents and full-stack apps you ship with a git push. That framing is also where the growth story lives: coverage around the raise put year-over-year revenue growth at more than 100%, a line that reads as "profitable, boring, safe to bet on."

If you are an investor, that headline answers your question: is this business compounding? If you are a tenant, it looks like it answers yours while saying nothing about your invoice. The timing is the tell — the raise closed in February, the repricing that funds the next year of that curve shipped in April.

What changed on April 23​

Render's old workspace model bundled two things: per-seat pricing and a generous flat bandwidth allowance of 100 GB on every tier. Most small teams never touched 100 GB, so the allowance functioned as a cap that rarely capped anyone. The April 23, 2026 "New Workspace Plans" unbundled both.

Here is the table from Render's own changelog entry and docs page:

Two changes point in opposite directions.

The win: flat workspace fees. A five-person team on old Pro at $25 per seat paid $125 in seat fees before compute; on the new model they pay $25 flat. For headcount-heavy teams, the seat savings beat the egress line at low traffic — Render's "better pricing for fast-growing teams" is honest about this part.

The cost: egress cut 20x on Hobby, 4x on Pro. The Hobby cut from 100 GB to 5 GB was noticed first — a SearXNG commit in April put it plainly: "Makes external search engines impractical without overage costs." The Pro cut to 25 GB looks smaller but lands harder on real apps; a single modest success crosses 25 GB without trying.

Every legacy workspace force-migrates by August 1, 2026 — no opt-out. If you are reading this after April, the bill below is already yours. Per-GB billing is fairer than the old 100 GB blocks, but fairer at the unit level does not change the slope of the total.

The math the headline hides: same stack, swept by bandwidth​

Take the stack most readers actually run: one web service, one background worker, one managed Postgres. Price it at the realistic floor for production — Standard-tier compute (1 CPU, 2 GB RAM) at $25 each, and Basic Postgres (1 GB RAM, 10 GB storage) at $20. Starter-tier sleeps and throttles, so it is not the comparison anyone stays on.

That gives a base before egress:

The base is fixed. What moves is the egress line. Because overage is a straight $0.15 per GB above 25 GB, the total is a line with a steep slope — exactly the shape that funding headlines never show.

Hobby at 5 GB is harsher, faster:

What the sweeps hide in a single "at 1 TB" anecdote:

• At low traffic, Render is competitive. At 25 GB, $95 for web+worker+Postgres+workspace is defensible — a five-person team's flat fee already beats old per-seat. Headcount-heavy teams win; bandwidth-heavy teams lose.

• At mid traffic, egress becomes the bill. At 500 GB, egress is 43% of the Pro invoice ($71.25 of $166.25); at 1 TB, 61%. Compute barely moves month to month; bandwidth decides the total, not the headline rate.

• Hetzner's line teaches the shape. A Hetzner CX22 — 2 vCPU, 4 GB RAM, 40 GB SSD, 20 TB included at €3.79/mo — cannot be bought per GB after that. EU overage past 20 TB is roughly €1/TB. For every workload in the sweeps above, Hetzner's egress column is zero with 19+ TB headroom. On Render the same compute costs differ by $300 a month depending on bytes shipped.

• Sensitivity is the message. 25 GB → 250 GB adds $33.75 on Render and nothing on Hetzner; 250 GB → 2 TB adds $269.70 on Render and still nothing on Hetzner. If you do not know your egress — and most teams do not until images, API JSON, or one viral post — you do not know your bill, and no valuation predicts it.

Why a valuation tells you nothing about next year's bill​

A business growing 100% year over year can be growing from two distributions that look identical in a press release:

Path A: more tenants at roughly the same bill. Ten thousand Hobby workspaces become twenty thousand, most under 25 GB, paying workspace fees and compute. Revenue doubles because customer count doubles. Your egress does not drive the aggregate.

Path B: same tenants at a higher bill because a meter now covers what was bundled. The same ten thousand workspaces each serve 40 GB more — normal once an app finds users — and 15 GB that used to sit inside the old 100 GB allowance is now billable at $0.15/GB. ARPU rises roughly $11 at 100 GB, $71 at 500 GB, $150 at 1 TB, with no new tenant. Revenue still doubles. The funding headline reads the same.

The two paths mean opposite things for you. Under Path A, your bill is decoupled from the company's growth. Under Path B, your bill is the growth — the same meter that turns "100%" into a $1.5B valuation turns your 500 GB month from $95 into $166.25. Headline and invoice are two faces of one repricing, not two signals where one validates the other.

An outside tenant cannot tell which path dominated from any number Render has published. Customer count versus ARPU, egress share of revenue — none is broken out, and private companies do not owe you that decomposition. So the valuation is not a safety signal. It is a claim about enterprise value whose inputs you cannot see, built on a repricing whose effect on your bill you can compute from the table above.

One more timing detail: legacy workspaces force-migrate by August 1. Until then, a chunk of the repricing's revenue step has not yet hit the trailing twelve months behind the "100% growth" number. Next year's growth will necessarily include a cohort that was previously sheltered. Whether that reads as "healthy compounding" or "same tenants paying more" depends on a number no press release puts in bold.

The self-hosted alternative and when it is not the answer​

Hetzner CX22 at €3.79 with 20 TB included is flat where Render slopes — even CX33 at €5.99 and CPX22 at €7.99 are an order of magnitude cheaper at 500 GB before ops time. But flat hardware is not free and not right for every workload.

When Render genuinely wins:

• Low-egress, collaboration-heavy teams. Internal tools or webhook-heavy backends under 25 GB pay only the flat $25 workspace for unlimited members. A five-person team saved $70–$120 migrating off per-seat; under 25 GB they never hit overage.

• Teams that cannot absorb ops. Managed Postgres at $20 that backs up itself, a git push that builds and rolls back without a MachineDeployment, a dashboard over git log rather than etcd — the sweep prices containers, not on-call.

• One tight burst. A marketing site spiking to 200 GB for a week then sleeping pays Render once. A Hetzner box must be sized for the peak even when idle unless you wire scale-to-zero; flat hardware overpays for bursty, low-average loads.

When flat hardware sharpens by comparison:

• Outbound bytes that scale with users — images, video, API JSON, docs served to crawlers all turn growth into egress. Metered egress taxes success.

• Predictable budgeting. A Hetzner invoice is the same number every month. Render's is a function of a variable your product team drives without seeing the meter.

• Valuation as platform-risk proxy. A $1.5B number does not answer "will my bill be the same next quarter?" A flat BOM does — you can print it.

The self-hosted counterpart is a Cluster API fleet on owned machines — declarative MachineDeployment, platform-owned node lifecycle, bandwidth already sunk into the hardware. You keep the git push while moving the egress column off the invoice: €1 per extra TB, not $0.15 per GB, and 20 TB per node is not a meter at all. That does not make self-hosting cheaper at every scale. It makes it constant where hosted egress is linear — and constant is what a budget needs when the vendor's own growth leans on the slope you pay.

What to do before August 1​

If you are on Render today, you need two numbers before August 1.

One: pull three months of bandwidth from Dashboard → Workspace → Billing → Bandwidth. Compare to your included tier — 5 GB Hobby, 25 GB Pro, 1 TB Scale — and multiply overage by $0.15. That is the delta migration adds, before any growth.

Two: run the same egress against the Hetzner you would actually provision — CX22 €3.79, CX33 €5.99, CPX22 €7.99, each with 20 TB, so usually zero. If the gap is small and you value managed Postgres and unlimited seats, staying is rational. If the gap is the largest line on the invoice, the valuation already explains why.

Growth headlines are for investors. Bills are for tenants. They stopped being the same number on April 23.

Render pricing details are from Render's April 23, 2026 changelog entry, New Workspace Plans docs, and Better pricing for fast-growing teams. Funding details are from the February 17, 2026 Business Wire release via VentureBeat, PYMNTS, and SiliconANGLE, led by Georgian at a $1.5B post-money valuation with $258M total raised. Hetzner CX22 pricing (2 vCPU, 4 GB, 20 TB, €3.79) is from Hetzner's 2026 Cloud pricing page; the per-TB overage past 20 TB is approximately €1/TB in EU/US regions.

Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own. Star the repo on GitHub or deploy your first app today. Flat hardware, no per-GB meter — the bandwidth is already in the box.