H1 2025 Recap: Key Signals from the Tech+Bio Frontier
A high-signal summary of some of the most important moves and trends shaping biotech and tech-enabled life sciences in the first half of 2025
Hi there! While hiding from the 38 °C heat under my airco—my best friend for this summer—I decided that it is the right time to reflect on an incredibly eventful but also a bit of a rollercoaster tech+bio space in the first half of 2025.
In this issue: Financial Climate & Deal-Making Power Shift — Gene Therapies are a Mixed Bag — AI Drug Discovery is Recalibrating — Big Data & Data Infrastructures Define Leaders — a Year of Next Generation Sequencing — Away from Animal Models — Brain-Computer Interfaces are Having Momentum
Before we dive in, I will be visiting the Aging Research & Drug Discovery Meeting - ARDD 2025, so let’s meet sometime between August 25-29 in Copenhagen!
Financial Climate & Deal-Making Power Shift
So, first things first, what does the market and venture capital space look like now when we enter H2?
I recently read a timely post by Audrey Greenberg, highlighting insights from reports by Endpoints News and DealForma. Biotech’s holding up thanks to big pharma – $98B in deals so far this year, with licensing surging (especially for Phase 2/3-ready programs) and blockbuster M&A like Merck-Verona and Sanofi-Blueprint leading the way.
On the flip side, venture funding has cratered, down 33% QoQ, with early rounds, IPOs, and follow-ons all at multi-year lows.
Also, human data is beating out platform hype, and in this shaky market, strategic M&A looks like biotech’s best path forward – especially with pharma still active, notably in China, which drove 38% of major licensing deals.
Speaking about the geography of biotech VC concentration, there is an interesting new biotech VC report from Nucleate Signal, which maps the top 10 biotech hubs by local investor count—led by San Francisco (1,752), followed by NYC (713), London (488), and Boston (478). Kudos to André Hurtado for pointing me to this post, btw.
The report introduces a new framework classifying cities by investor density × capital retention, grouping them into four types like Resilient Engines (e.g., SF, Boston) and Capital Exporters (e.g., NYC, London). Notably, San Diego, often considered a strong VC hub, is labeled Externally Dependent, raising questions about local reinvestment. The report, "Built from Within", was written by Lauren Stanwicks and Pranita Atri and could be interesting for those of you thinking about biotech investment strategy in H2 and beyond.
Finally, China… of course.
China’s biotech scene made some serious waves in H1 2025, and the industry’s paying close attention. Big pharma didn’t hold back; 14 major licensing deals were signed with Chinese firms, totaling over $18 billion, up from just two last year. Suddenly, a third of all deal value globally is coming from China-based innovation!
Investors are jumping in too: the Hang Seng Biotech Index is up nearly 80%, and 34 Chinese biotechs filed for Hong Kong IPOs in just the first half of the year.
Companies like Akeso and XtalPi made headlines in H1, for instance, with competitive drugs and AI-powered R&D, showing China’s not just catching up, but possibly also setting the pace in some areas.
That being said, there’s a real geopolitical edge to all this. U.S. policymakers are probably starting to worry about biosecurity and data-sharing, so while the business side sees opportunity, there’s growing tension in the background. Bottom line: China’s biotech rise is real, fast, and making the rest of the world take notice... in Europe too.
Gene Therapies, a Mixed Bag
The gene-editing space in H1 2025 was a mix of excitement and caution. On one hand, big players like Eli Lilly doubled down, buying Verve Therapeutics in a $1.3 billion deal after promising CRISPR-based cholesterol data. The global market’s still growing fast, worth around $4.5 billion now and climbing at a 10–14% annual rate as per Bloomberg Law.
But funding has cooled off sharply (as per DealForma/Reuters), dropping from over $8 billion in 2021 to just $1.4 billion in 2024, forcing companies like Editas and Prime Medicine to trim pipelines and cut costs.
At the same time, research progress is real: from the first successful prime editing treatment in CGD, to a baby getting a custom LNP-based CRISPR therapy.
Still, safety concerns loomed large… Sarepta encountered obstacles after multiple patient deaths and FDA intervention, possibly shaking investor confidence in AAV-based delivery.
Overall, it’s a moment of recalibration, so to speak… Plenty of long-term optimism, but with more scrutiny, tighter budgets, and a sharper focus on what actually works.
Now, speaking about the future of gene therapies, I think it will be increasingly shaped by the emerging technologies, like LLMs and other modeling tools. Gene editing is extremely complex and multifaceted, so it is a ripe area for AI models to shine, at least for those that can generalize across datasets and contexts.
Especially interesting to me is a new powerful modeling tool, just published in Nature Biomedical Engineering, which is called CRISPR-GPT. It is an autonomous, language model–driven agent designed to plan and execute CRISPR-based gene-editing experiments across modalities like knockout, base editing, prime editing, and epigenetic editing.
Conceptually, it integrates reasoning, task decomposition, and domain-specific tools to automate the entire experimental pipeline—from guide RNA design to protocol generation and data interpretation. Its architecture uses modular agents to coordinate planning, execution, and user interaction, enabling flexible operation modes (e.g., full automation, guided workflows, or targeted Q&A). Technical tools are invoked as needed, but the core innovation lies in treating biological experimentation as a structured, language-guided problem solvable through iterative agentic reasoning.
AI Drug Discovery: Industry Recalibration
So, as we already mentioned, CRISPR-GPT, let's continue with this topic.
The first half of 2025 in AI drug discovery has been fast-moving but mixed — packed with new launches, pilots, and tech releases, yet still light on hard validation.
Companies are rolling out increasingly sophisticated platforms: Genentech introduced SpatialAgent, an LLM-based system for spatial genomics; Harvard and Roche launched COMPASS, a biologically interpretable AI that beat traditional biomarkers in predicting immunotherapy response; and NVIDIA’s Evo 2, trained on 9 trillion nucleotides, became publicly available through BioNeMo, promising new capabilities in genome and protein modeling.
Investor interest remains strong, but the funding landscape is shifting. Capital is concentrating around established platforms and later-stage AI-biotech hybrids like Isomorphic Labs, which secured $600M to scale its unified drug design engine.
Meanwhile, new entrants are finding it tougher to raise, in my opinion, unless they bring proprietary data or visible pipelines — a clear pivot from the broader enthusiasm that characterized 2021–2023. There are outliers, though, like Sam Altman's Retro Biosciences that eyed a $1B round to extend lifespan via aiming its foundation model at cellular reprogramming, which is a new venture without much to show yet. But I think those aren’t really reflective of the more general VC trend of dumping hype and focusing on practical ROI.
Speaking about big pharma, CB Insights releases 2025 Pharma AI Readiness Index—the report ranks Eli Lilly, Merck KGaA, and Bayer as the top three among 50 pharma companies based on AI execution, innovation, and talent.
While I won’t speculate about the report’s methodology, a quick glance at recent news brings up, for instance, that Eli Lilly partnered with Creyon Bio in a deal worth over $1 billion to co-develop AI- and quantum chemistry–designed RNA-targeted oligonucleotide therapies, granting Lilly exclusive rights to lead candidates across multiple disease areas. Merck KGaA adopted Quris-AI’s Bio-AI clinical prediction platform to assess preclinical small molecule candidates, following a two-year validation study showing superior accuracy in predicting drug-induced liver injury compared to traditional safety models.
Finally, Bayer entered a multiyear collaboration with ConcertAI to access its Translational360 platform—an AI-powered molecular database built on 9 million U.S. cancer patient records—to accelerate precision oncology drug development and trial design.
But for all the activity, clinical validation remains limited. While a few companies, like Hong Kong-based Insilico Medicine, with its Phase II assets and multiple Phase 1 assets developed arguably faster compared to classical non-AI routes, have shown the potential of AI-native pipelines, most AI-designed compounds are still stuck in preclinical or early clinical stages.
There is a BiopharmaTrend’s report I would mention here that provides a pretty deep analysis of the current state of the AI-inspired clinical candidates by various companies. The sector has yet to produce a fully AI-developed drug that reaches market approval (if “fully AI-developed drug” is even possible), and timelines for clinical impact continue to stretch. The bold claims from just a few years ago — that AI would compress discovery into months and de-risk failures — are now being met with a more tempered, critical lens.
Big Data and Data Infrastructures
While the progress in AI is ongoing and promising, the key enabler of AI transformation is, of course, data. At the beginning of 2025, I wrote a post where I predicted that big data partnerships would be a new trend in 2025.
Well, it seems data plays are indeed expanding: HuBMAP’s Human Reference Atlas now maps nearly 1,200 cell types across 65 organs, and Truveta, backed also by Regeneron and Illumina, raised $320M to build the largest exome-linked clinical database in the U.S.
The Human Cell Atlas, built by 3,600 scientists from 102 countries, maps 62 million cells across human development, offering a detailed, AI-powered blueprint of the body. It is probably one of the largest data projects in biotech I’ve seen so far.
Among some notable data-centered deals, Truveta has secured a $320 million investment from Regeneron, Illumina, and 17 U.S. health systems to build the world’s largest genetic database, targeting 10 million sequences, by linking genomic data with anonymized health records to accelerate drug discovery and personalize care.
Another vivid example is Tempus AI, a company that aggregated a lot of research and clinical-grade data, and built infrastructure. In 2025, Tempus AI signed a $200 million deal with AstraZeneca and Pathos AI, who tapped Tempus primarily for its unique role as a source of deeply structured, multimodal patient data to build cancer foundation models. Similarly, Boehringer Ingelheim and Illumina partnered with Tempus to gain access to its massive real-world clinical and genomic datasets, underscoring that Tempus’s value lies not just in AI tools, but in being a premium data provider fueling precision medicine.
Next Generation Sequencing
Genomics sequencing is a 'hot' topic this year. Roche is back on track with the introduction of Sequencing by Expansion (SBX), a next-generation sequencing (NGS) technology that converts DNA into Xpandomers, molecules 50 times longer than original DNA to enhance signal clarity, speed, and scalability.
Led by Mark Kokoris, SBX offers high accuracy (99.80% for SNVs, 99.48% for InDels), ultra-fast workflows (blood sample to VCF in under 8 hours), and high throughput (7 human genomes at 30X per hour), with early access in 2025 and full commercialization in 2026.
In another big milestone, Ultima Genomics launched the UG 100 Solaris sequencing platform at AGBT 2025, featuring upgraded chemistry, software, and workflows that boost sequencing output by 50% to 10–12 billion reads per wafer while cutting costs by 20% to $0.24 per million reads, advancing the company toward the "$80 genome". The SNVQ60 ppmSeq mode enables 30X genome coverage with just 2 nanograms of DNA, and the early-access UG 100 Solaris Boost mode supports up to 100 billion reads per day, with full release expected in late 2025.
John Overton (Regeneron Genetics Center) emphasized its role in the UK Biobank Pharma Proteomics Project, one of the world’s largest proteomics datasets, and Johnny Yu (Vevo Therapeutics) commented on the use of UG 100 to create a 100-million-cell transcriptomic dataset supporting AI-driven drug discovery, which will be open-sourced with NVIDIA's backing. CEO Gilad Almogy described Solaris Boost as part of Ultima’s mission to scale genomic data generation for research and clinical applications.
Away from Animal Models
As you probably know, in a major policy shift, the FDA and NIH are accelerating the move away from animal testing in drug development, endorsing New Approach Methods (NAMs) like organoids, organ-on-a-chip (OOC) systems, and AI-driven toxicity prediction models.
While we are in the early days of being able to really substitute animals with alternative models, in my opinion, the business opportunity is real and prominent.
There is a wave of companies, including Axiom Bio, with a deep-learning model trained on 9 billion liver mitochondria images; Certara, integrating PBPK/QSP modeling with AI via its Non-Animal Navigator and Simcyp platforms; VivoSim Labs (formerly Organovo), offering 3D bioprinted organoids and the NAMkind liver toxicity platform; Toxometris.ai, providing ensemble models for over 40 ADMET endpoints and carcinogenic risk scoring; and so on.
In the organ-on-a-chip space, notable players include Emulate, Tissue Dynamics (fusing OOC with real-time metabolic monitoring and AI), InSphero, Mimetas, Quris-AI, 4DCell, AlveoliX, TissUse (with multi-organ microfluidic circuits), Dynamic42, and academic labs building human assembloid models like Stanford’s hASA.
These tools are increasingly replacing in vivo assays by offering functional tissue behavior, immune integration, and dynamic drug response modeling.
Brain-computer Interfaces are Having Momentum
Finally, my new favorite topic, BCIs, is a wild tech+bio area to watch. In H1 2025, brain-computer interfaces are leaping out of labs and into real-world use. Obvious headliner is Neuralink, which has let paralyzed individuals control cursors, type emails, and even play chess using just their thoughts, while still navigating challenges like long-term signal retention, you all heard about it. Do you think Neuralink is the only big deal in BCI space? Think twice…
Here is a category-defining Synchron’s Stentrode, a minimally invasive BCI threaded through the jugular vein, is already in human trials and lets users with paralysis control apps, smart home devices, and even voice assistants with thought—backed by Nvidia AI, Apple Vision Pro, and partners like Amazon and Apple.
Another neurotech startup Paradromics successfully implanted its Connexus brain-computer interface into a human for the first time—demonstrating safe implantation and neural recording in a clinical setting.
Just this April, the FDA granted human investigational device exemption (IDE) to Precision Neuroscience, allowing the company to begin clinical trials of its minimally invasive robotically delivered brain implant, positioning it as a direct rival to Neuralink.
On the research side, Harvard’s team created ultra-soft, stretchable implants inserted at the embryonic neural plate stage in tadpoles. These tiny fluorinated-elastomer devices record single-neuron activity without harming development, offering a glimpse into brain growth and next-gen BCI materials.
Finally, global dynamics are shaping the future of BCIs, and I can detect the beginnings of the new geopolitical rivalry, similar to what we observe in the gen AI space. For instance, among a wave of BCI-related news from China in H1 2025, the latest is that China has launched its first clinical trial of a BCI, enabling a tetraplegic patient to control digital devices with thought alone just weeks after implantation, marking a significant national milestone and positioning China as the second country globally, after the U.S., to reach this clinical stage...
In short, BCIs have crossed a major threshold; the technology is finally getting into the clinical realm. This will inevitably raise many questions beyond just performance and stability: ethics, privacy, even philosophical questions, it will be exciting to watch…
Anyway, let me know what you think! Any other signals this year that may be crucial for strategy and decision making? Share your thoughts below…