How much should it actually cost to manufacture an implantable, biologics-based artificial organ? People reflexively ask when such a technology will arrive, but far fewer ask how much it should cost. That is the more uncomfortable question, and probably the more useful one. The silence likely comes from two assumptions: either the answer is so expensive that it feels irrelevant to ordinary people, or the task is technically impossible.

Fortunately, neither assumption is quite right. Tissue engineers and startup founders would reasonably argue that biologics-based artificial organs are not forbidden by first principles. No one has shown that an implantable artificial organ violates the laws of physics; the issue is cost, complexity, and time.

If there is a price tag, then how high might it be?

Because there has been far less progress and investment in whole-organ regeneration than in frontier AI, AGI is a useful proxy. Both AGI and artificial organs sit in a category of technologies people describe in near-mythic terms: they tackle unusually difficult problems, require enormous infrastructure, and promise effects that extend beyond engineering into economics and social organization.

(Subscribe to “The Lattice Podcast” for bonus episodes here.)

How much to AGI?

AGI is easier to use as a benchmark because there are already large, publicly available estimates. The five largest US cloud and AI infrastructure providers, Microsoft, Alphabet, Amazon, Meta, and Oracle, are projected to spend roughly $660–$690 billion on capital expenditure in 2026, nearly double 2025 levels (Futurum Group). Project Stargate has been described as a $500 billion effort targeting 10 gigawatts of AI infrastructure by 2029, with tens of billions in chip purchases alone before replacement cycles and upgrades (Stargate LLC).

Longer-range projections are even larger. Goldman Sachs has estimated about $7.6 trillion in AI infrastructure capital expenditure between 2026 and 2031 across chips, data centers, and power systems (Goldman Sachs via Economic Times). McKinsey forecasts 156 GW of AI-related data center capacity demand by 2030, which it estimates will require around $5.2 trillion in capital expenditure, with annual infrastructure funding needs exceeding $1.4 trillion by the end of the decade. (McKinsey)

The cost of operating a frontier AI lab is already measured in billions. Epoch AI estimated Anthropic spent roughly $9.7 billion in 2025, including about $6.8 billion on compute alone, with compute accounting for 57–70% of total spending at frontier AI companies. Even if training costs fall over time, frontier AI remains a deeply capital-intensive activity. (Epoch AI)

This spending is beginning to matter at the macroeconomic level. In Q1 2026, AI-sector capital expenditure reportedly reached about $174 billion, a 72.8% increase over the prior year, accounting for a significant share of US GDP expansion that quarter. Wells Fargo projections suggest this will rise to more than 3% of US GDP by the end of 2026, surpassing the investment share seen at the peak of the 1850s railroad boom.

This is just to reach AGI, potentially. These figures do not include the ongoing cost of inference and the energy constraints data centers face, which add a large recurring layer of spending.

There is also a harder accounting problem. There is no reliable estimate of how many dollars must be spent to generate one dollar of revenue at frontier AI companies, but it is almost certainly more than one. Most consumer AI use is heavily subsidized.

That said, revenue is not the same as usefulness. AI systems have improved remarkably, but they still make basic and sometimes costly errors. The progress has been real; the reliability has not.

How much to implantable artificial organs?

Artificial organs are not one category. Some will be far easier and cheaper than others. Products closest to commercial reality tend to be avascular or structurally simple tissues rather than dense, highly vascularized solid organs.

That pattern is already visible. Organogenesis has commercialized cell-based skin products. Nanochon is pursuing an acellular 3D-printed cartilage implant in clinical trials. Hemocyte is developing acellular vascular grafts, with ambitions to expand into other organ systems. In investment terms, the case for bladder, trachea, cartilage, or small vessels is far cleaner than the case for kidney or heart: lower manufacturing complexity, shorter timelines, a more navigable regulatory process, and clearer reimbursement pathways.

Solid organs such as the kidney, liver, and heart are different. They combine dense vascularization, multiple interacting cell types, demanding microarchitecture, and severe functional requirements. Among tissue-engineering targets, they remain the hardest and are widely described as the field’s holy grail.

Why the kidney is the right example

The kidney is a useful case study because the medical need is obvious and the economics are unusually stark. A kidney transplant in the US costs about $442,500, and recipients generally require lifelong immunosuppression while graft survival often falls in the 15–23 year range. Dialysis, the main alternative for end-stage kidney disease, costs roughly $90,000–$100,000 per patient per year in the US (NIH).

A Swedish real-world study found that kidney transplantation avoided 66–79% of expected healthcare costs over ten years compared with dialysis, corresponding to roughly €380,000 in savings per transplanted patient. For a patient who remains on dialysis for a decade while waiting for a transplant, direct treatment cost alone approaches $900,000 to $1 million before complications, hospitalizations, or lost productivity are counted.

At the system level, renal replacement therapy consumes a striking share of healthcare spending. Estimates suggest governments spend roughly 3–5% of annual healthcare budgets on renal replacement therapies, while the global ESKD population may reach 14.5 million by 2030 (PubMed Central). That makes kidney failure not just a scientific problem, but a very large and recurring budget problem.

What is being spent now

The ARPA-H PRINT program focuses on the kidney, heart, and liver because the kidney is the most transplanted organ and has the longest wait lists. There are roughly 120,000 people on the US organ waiting lists, while only about 45,000 transplants are performed each year. The 75,000-person annual kidney shortfall, at $90,000–$100,000 per year in dialysis costs, represents roughly $6.75 billion in excess dialysis spending that a scalable bioprinted kidney could theoretically displace.

Against that backdrop, the scale mismatch with AI spending is hard to ignore. OpenAI reportedly spent around $22 billion in 2025 to generate $13 billion in revenue, while the entire ARPA-H PRINT program, the most ambitious US government effort focused on transplantable organs, was funded at $65 million. The disparity between what is spent to generate AI responses versus what is spent to solve organ failure is more than 330:1!

The contrast becomes even more striking when set against the broader landscape of healthcare and regenerative medicine. US national health expenditure reached $5.3 trillion in 2024, equivalent to 18.0% of GDP, and CMS projections suggest it will approach 20.3% of GDP by 2033. The US spends more on healthcare as a share of its economy than any peer nation, and by a wide margin: the OECD average among high-income countries was around 11.2% of GDP in 2024 (OECD Health at a Glance 2025). Yet the regenerative medicine sector, the field most likely to structurally reduce that burden, is estimated at roughly $45 billion globally in 2025, growing toward $97 billion by 2035 (Market Research Future). That is less than 1% of US healthcare spending in a single year. NIH’s total extramural grant investment was about $35.3 billion in FY2025 across all of biomedical research (NIH), while ARPA-H’s most ambitious organ regeneration program was funded at $65 million. The spending priorities embedded in those numbers are worth examining.

Should more money go into tissue-engineered kidneys?

Maybe, but not automatically.

Development costs for a tissue-engineered or bioprinted kidney would likely reach the high hundreds of millions to multiple billions of dollars. Vascularization, maturation, integration, reproducibility, and manufacturing all remain unsolved at a clinically useful scale. Kidney bioprinting efforts are currently funded in the tens of millions per project, for example, an ARPA-H-supported program of up to $24.8 million over five years, while reviews continue to describe the field as preclinical. Bioprinters may range from roughly $5,000 to $400,000 each, but the dominant costs lie in cell manufacturing, bioinks, perfusion systems, quality control, animal studies, GMP-scale-up, surgery, and regulatory work.

The final per-patient price of a first-generation tissue-engineered kidney will almost certainly be high at launch, likely in the high six figures and possibly above today’s transplant cost, even if the economics improve with scale. One industry analysis argues that a 3D-printed organ could eventually compete with transplant costs and long-term care, but this remains speculative because no commercial bioprinted kidney exists yet. A useful analogy comes from stem cell–based tissue-engineered airway transplants: manufacturing costs were estimated at roughly $12,690–$27,490 per graft, but total patient costs ranged from about $133,140–$144,180 in modeled future trials and reached $553,140 or more in early real-world cases based on a 2015 paper. The fabrication step was only one part of the total bill.

The harder question

The parallel with AGI suggests a broader choice. Society may benefit more, at least for a time, by not pushing every ambitious technology to its physical limit as quickly as possible. Artificial organs are not impossible; it may still be rational to delay the most ambitious versions if the same capital could fund cheaper, scalable interventions that help far more patients in the near term. For example, drugs or devices that would decrease the total number of patients needing transplants.

That trade-off is uncomfortable but real. It may mean getting a fully tissue-engineered kidney decades later than is theoretically possible. It may also mean saving more lives sooner by reducing organ injury, offering alternative treatments like dialysis, improving transplantation logistics, developing simpler engineered tissues, and deploying other practical technologies that cost less and reach more patients faster.

I think the AI industry should consider such alternatives seriously as well to solve society’s immediate problems at scale, rather than AGI.