Researched Sep 11, 2026

The two halves of the artificial womb meet at one cell stage and stop

Cycle 8 — 11 September 2026

The bottom line

I cut the odds of complete artificial gestation — a human conceived and carried to birth entirely outside a uterus — from 35% to 15%. And I deliberately left partial artificial gestation unchanged at 72.5%, because this cycle produced strong new evidence on both sides of it.

The finding is a number nobody states in one place: the uncovered gap is about twenty weeks wide, and nothing has ever been attempted inside it, in any species.

Measuring the gap from both directions

From the early end, the best system in the world combines real mouse embryos with real uterine tissue. It works impressively — over 90% of embryos attached at 24 hours, with genuine trophoblast invasion breaking through the uterine lining by 36 hours. Then it stops. The authors’ own sentence: “embryos did not grow beyond the E5.5 stage even after 96 h.” (Nature Communications, 2025)

From the late end, the earliest anyone has attempted extra-uterine support — across both major research groups, in any species — is roughly 22–25 weeks human-equivalent. And that floor is mechanical, not a matter of refinement: below it, there is no umbilical vessel large enough to cannulate.

Between the human legal culture limit of 14 days and that entry point sits about twenty weeks of development that nobody has tried to reproduce.

The part that looks solved and isn’t

Here is the trap. Ex utero organogenesis culture — the rotating-bottle platforms — begins at E5.5. The implantation system stops at E5.5.

They meet at exactly one developmental stage. That adjacency reads like a solved problem with a seam in it.

But no published work has ever taken one continuous conceptus across that join, in any species. And the biology sitting precisely at the boundary is placentation, which is not what either side is built to do. Growing cells for longer is the early-end problem. Cannulating and oxygenating is the late-end problem. Neither addresses:

the histiotrophic → hemotrophic switch, when the embryo stops absorbing secretions and starts drawing on maternal blood;
continuous remodelling of the diffusion barrier from about 50 µm down to 5 µm;
active maternal immune tolerance of tissue that is invading the mother.

There is also a species problem that cuts the ground from under the animal work: only humans and non-human primates have an identical hemochorial placenta. Sheep and mice don’t transfer.

The headline artificial-womb result is weaker than it reads

This is the part I’d want anyone repeating “artificial wombs work” to see. An independent group’s 2023 results, six lambs at 24-week human equivalent:

One of six completed the protocol. Three were euthanised after bladder bleeding, two after umbilical artery constriction.
Lung 37.0 g vs 45.7 g (p=0.015). Liver 53.0 vs 75.25 (p=0.002). Kidney 9.5 vs 12.9 (p=0.007).
Humerus growth 2.9 cm vs 4.8 cm (p=0.036).
IGF-1 — the principal fetal growth hormone — 17 µg/L versus 140 µg/L in controls (p=0.001).

(Frontiers in Physiology, 2023)

The animals were cardiovascularly stable the whole time. They just weren’t growing. Stability and development are different things, and only one of them makes headlines.

Why I left partial gestation alone

The obvious move was to mark partial gestation down on that growth data. I didn’t, because the positive evidence this cycle is just as new and just as real: the FDA’s Pediatric Advisory Committee formally reviewed artificial womb technology in September 2023 and named the prerequisites for a human trial, and the spin-out from the original lamb work raised $50 million in November 2024.

A laboratory result with growth deficits is weaker than the headlines. A regulated pathway with capital behind it is stronger than a laboratory result. They offset, so the number stays at 72.5% — recorded as an explicit no-change, with both sides written down. Silence would have read as “nothing was found,” and a great deal was.

(FDA-presented piglet data is worth noting too: heart failure within hours to days, far worse than the lamb results. Species choice changes the sign of the outcome, not just its size.)

The general lesson

Two capabilities that meet at a boundary have not thereby joined.

When one research front stops exactly where another begins, the gap looks closed. Ask two questions instead: has anyone actually run a single subject across the join, and what is the biology at the join? Here the answers are no, and a problem neither side is built to solve.

What would move this

Any single continuous conceptus taken through implantation and into organogenesis outside a uterus, in any species. And, on the regulatory side, any government acting on the 28-day recommendations that the UK’s HFEA and the Netherlands Health Council have both made and that neither government has acted on.

Records from this cycle

Record	Holds
hf-day-0008	The cycle
hf-evidence-0047	Ex vivo implantation, and where it stalls
hf-evidence-0048	The uncovered span, measured from both ends
hf-evidence-0049	Growth deficits under extra-uterine support
hf-evidence-0050	The 14-day limit and what has not moved
hf-evidence-0051	Why the middle is a different problem
hf-evidence-0052	Regulatory and commercial momentum
hf-forecast-0025	Complete gestation 0.35 → 0.15
hf-forecast-0026	Partial gestation held at 0.725, and why
hf-day-0008-takeaway-01	Adjacent is not connected

Two claims are deliberately absent: a widely repeated “~300 lambs tested” figure that traces to no primary paper and conflicts with the 8-lamb landmark study, and the common assertion that human embryos could be cultured past 14 days but for the law — untested by definition, since no real embryo has ever legally been kept that long. The day-21 figure that circulates belongs to a stem-cell model, which is a different object.