"Where's the gold?" — A Gene Therapist's Tale of Cell and Gene Therapy

We were promised gold. We made & bought our shovels, rode the bull, and arrived at the gold fields of Northern California. But all we saw was a wasteland.

This post is about the rise, fall, and inevitable resurgence of cell and gene therapy (CGT)—told by me, a rando on the internet who’s spent the past ~decade in the field.

Luxturna was a success. Humankind gave back light to a thirteen year old boy!

Luxturna was an Adeno Associated Virus (AAV)-based gene therapy to treat rare genetic inherited retinal disease. Giving back sight to children and adult patients, it showed the world the inevitable futures we were about to live in. It signified a new era of medicine, and an abundant world where no disease would be “untreatable.”

It amplified the wave of investment and research that was already started in 2013 with the invention of CRISPR. Two widely publicized breakthroughs in one field within 5 years is exceptionally rare in science. And we were all excited. CRISPR/Cas9 was the first iteration, but countless other variants like CasX and Cas13 were being developed, with high-signal companies from UC Berkeley and MIT/Harvard like Caribou to Sherlock being spun out with pomp and circumstance.

On the cell therapy side, the momentum was just as strong. Kymriah and Yescarta—the first CAR-T therapies—had been approved in 2017, and Novartis and Gilead were scaling manufacturing to meet demand for what looked like a new standard of care in blood cancers.

“Precision medicine” was the catch phrase.

With tools like CRISPR and delivery methods like AAV now de-risked in clinical trials, we could now plausibly engineer precise, long-acting medicines that could treat, and potentially cure, any disease, in one shot.

2018 to ~2022 were some of the most explosive years for CGT, with money flowing across the field to prepare for the gold rush. Investments covered a few key areas that needed improvement: cargo R&D (CRISPR, siRNA, CAR-T), delivery vehicle engineering (AAV, LNP, non-viral capsids), and manufacturing (production, QC).

In a gold rush, you make your money by selling shovels. And that’s exactly what the industry did. Across these ~three areas, billions of dollars in investments, partnerships, and acquisitions flowed that told the world:

“The Gold Rush is Here”

2022 was a great year to be in CGT. Spark, the company that made Luxturna, was so hot, scientists were being poached from there like OpenAI taking AI scientists from Meta. Forge was making it big. Kriya was coming up, and companies were starting to make big, seemingly generational plays.

CAR-T was already proving itself commercially. Yescarta and Kymriah were generating billions in cumulative sales, and Carvykti’s approval for multiple myeloma in 2022 expanded the addressable market beyond lymphomas and leukemias.

Could we custom-design, engineer and sell AAVs, the main delivery vehicle of gene therapies? How could gene therapy cure common disease, after curing rare disease?

These companies, if successful, could capture tens of billions of value in the gene therapy market. The +$400k price tag of Luxturna (per eye) doesn’t even matter if we can calculate the net present value of alternative therapeutics on the market if the gene therapy is a one-shot treatment. The excitement of CGTs during that time was backed by real optimism of what the modality could do for patients.

Yes, the treatments were expensive. But it’s fine if the treatment delivered better value than the existing alternatives in a one-shot format.

Yes, immunogenicity was a problem. But only sometimes when the dose was too high.

Yes, manufacturing was hard. But scale would fix that.

It felt as if every hard problem had a “but” that made it manageable. Every risk had a hedge. The field had built such momentum that the obstacles felt like simple engineering problems that we’ll solve for in time. The money was flowing. And the patients who could see again were the proof.

I think the market, and the field as a whole, matured.

What matured means, practically, is that the heuristic-level optimism from 2018–2022 met the harsh reality of biology and sometimes plain physics.

Gene therapy wasn’t a one-time, lasting cure. Luxturna didn’t restore 100% vision. In some patients, the effects wore off significantly. People in the field intuitively thought this, but there was a gradual shift in the public rhetoric from “cure” to “treatment.” And with that shift, the entire cost-benefit calculus changed. A $2M+ price tag for a cure was defensible, but for a treatment that might wear off in a few years? That’s a much harder sell.

Gene therapy was finicky, and sometimes deadly. Neutralising antibodies reduced therapeutic effects in patients who’d been previously exposed to AAV—which, depending on the (sero)type of AAV, could be a significant fraction of the population. Worse, some patients died. And they were usually still children. Insufficient neutralising antibody screening on patients, excessively high drug doses to compensate, and the utter gut-punch of a little boy dying from drug-related liver damage weighed very heavily on the field.

Manufacturing and QC remained genuinely hard. AAV capsids assemble well, but the genome replication process is error-prone—snap-backs, truncations, and unwanted DNA species get packaged alongside the intended cargo. Separating these differently-sized molecules at scale, consistently, with the purity regulators require, pushed the limits of existing bioprocessing infrastructure.

The infrastructure to deliver such a fragile drug didn’t exist. The Moderna vaccine, an mRNA encased in a lipid nano-particle (LNP), is notoriously difficult to distribute and requires ultra-cold chain storage. Gene therapies face a version of the same problem, often with even more stringent handling requirements. Patients became the last frontier, something small molecules and other modalities solved for ages ago. The manufacturing and infrastructure problems are both what ARPA-H has recently addressed in their new GIVE program.

And even when gene therapies worked, the economics didn’t. Roche acquired Spark for $4.3 billion in 2019, largely on the strength of Luxturna and a hemophilia A pipeline. By 2024, Luxturna was generating roughly $20 million a year in sales—down 59% year-over-year—and Roche took a $2.4 billion impairment on the unit. Bluebird bio, once valued at nearly $10 billion, was sold to private equity for $29 million after failing to commercialise three approved gene therapies. Pfizer’s hemophilia B therapy Beqvez was discontinued less than a year after approval; reportedly, zero patients had been treated. The payer infrastructure for million-dollar one-shot therapies never materialised at scale.

With this, the market makers started to step back.

Novartis confirmed a phased wind-down of a San Diego technical R&D site for gene therapy (July 2024). Pfizer discontinued commercial availability and global development of its hemophilia B gene therapy Beqvez (February 2025). Roche restructured the very company that made Luxturna, Spark, shelving the hemophilia A program (restructure in 2025). Vertex stopped using AAV as a delivery mechanism entirely (May 2025). Biogen discontinued all AAV gene therapy programs and restructured the team (September 2025). Takeda exited cell therapy research (October 2025). Novo Nordisk halted cell therapy R&D (October 2025).

Seven major pharma companies. Eighteen months.

It had been about four years since those new CGT startups were founded. They matured, developed targets, and built out their pipelines, and prepared pre-IND packages expecting that someone would be there on the other side to receive it.

The licensing partners were gone. The market-shapers who set the terms for every deal in this space were suddenly, collectively, disinterested. In that moment, an entire generation of CGT companies had to think critically about what part of their business model still actually worked.

It’s been a tough two years for small-cap and pre-Series B biotech startups. Many of these companies are still around today. Some pivoted targets or modalities. Others are running lean, extending runway, waiting for the licensing market to recover. A few didn’t make it.

The punch down of the platform biotech has been brutal, and hard to evade for many CGT companies founded between 2018-2021. But there’s still much to learn, and the field’s story is far from over.

Firstly, we continue to cure previously incurable disease with CGT. Just last summer, we were able to create and administer a life-saving gene therapy into a 6-month-old baby, KJ, going from identifying the genetic mutation to a treatment in just six months total. A stunning and herculean effort by the team at UPenn. Casgevy showed what’s possible for ex vivo cell therapies: CRISPR-edited autologous stem cell therapies that enable independence from routine transfusions for many with beta thalassemia.

More broadly, CAR-T has become the commercial bright spot of the CGT field, the market reaching roughly $5 billion in 2024—though uptake for newer ex vivo therapies have been slower than expected, echoing the same payer access challenges that plagued gene therapy.

The science never failed, just our heuristic-level expectations. We expected one-shot cures from first-generation tools, and when we didn’t get them, we called it a failure of the modality rather than a failure of our own timelines and assumptions. The fundamental insight—that you can deliver genetic instructions to fix broken biology—remains as true as it was in 2018, and it continues to give patients a chance at a better life.

And all the diseases we said gene therapy would cure? They haven’t been treated by anything else since.

CGTs have undoubtedly transformed rare disease care, and the broader field still has much to learn from them. It tells us where durability breaks down, where immune responses escalate, and how to justify small N manufacturing runs. Transferable learnings are rare, especially in clinical trials where the information is siloed, but thankfully there are great people working on that too.

Finally, gene therapy is still a powerful modality for discovery. AAV is still the only way to get cargo into very specific cell types in hard-to-reach areas like the brain. ARIA’s neuro programs and even the new Merge Labs heavily rely on AAV-based approaches for genetically interfacing with the brain. AAVs are also now the best and easiest way to genetically modify new animal species—opening up entirely new research models that weren’t possible five years ago.

If there is one, it should be for AI-forward bio companies, which are now reaping the benefits of the hype cycle we’re in today.

High valuations, high expectations, and a fancy model are appealing to the glossy-eyed “deeptech investor.” But you’re rarely re-inventing the wheel, especially the exit strategy, if you’re a biotech play at the end. If the end-buyers for ~traditional therapeutics are the same, you’re still subject to the whims of pharma and the market makers.

CGT companies thought they had a moat because their tech was novel. AI-bio companies think they have a moat because their model is novel. But pharma doesn’t acquire moats—they acquire assets. And if the cost of the asset is all that matters, capital-intensive model-forward companies may find the multiples aren’t where they thought they were.

History has a tendency to repeats itself. And humans, ultimately, are too often emotionally persuaded by hype and hot new technologies, ignoring the boring and menial work to actually get there.

Patients are the final frontier, and more technologists have to appreciate what this truly means. Today’s AI for Science wave, if we’re not careful, might end as a mere hype cycle if we continue ignoring the real, boring, hard work to get us there.

Conflicts: I’m a shareholder in several gene therapy companies & startups.

Also: This is the first and last time I’ll unironically call myself a “gene therapist.”