Merck × Google Shock | The Full Picture of a ¥100 Billion AI Pharma Revolution
機械翻訳 / Machine-translated
A
AI Friends
2026年4月26日・39分で読める
機械翻訳 / Machine-translated
@aifriends
AI Friends（https://aifriends.jp）のクロスポスト公式アカウント。AIツールの紹介・使い方・できることを、中学生でもわかるやさしい日本語で届けます。
"It used to take 10 years and hundreds of billions of yen to bring a single new drug to market" — that conventional wisdom is beginning to crumble.
On April 22, 2026, U.S. pharmaceutical giant Merck (known as MSD in Japan) announced a multi-year agreement with Google Cloud worth up to $1 billion (~¥150 billion).
This is one of the most sweeping decisions in pharmaceutical history: AI embedded into every business function across the company, from drug discovery to factory operations to sales.
"Why now?" "Does this affect medicines in Japan?" "How will our hospital visits change?" — let's unpack what this announcement really means, in language anyone can understand.
Let's start by laying out the basics in five minutes.
The announcement was made on April 22, 2026, during the keynote of Google Cloud's annual event, "Cloud Next 2026," held in Las Vegas.
Cloud Next is the once-a-year stage where Google shows the world what it's truly serious about in its cloud business — and the star of the show was Merck × Google Cloud's company-wide AI contract worth over ¥100 billion.
Think of it like a baseball team giving a prized rookie the starting spot on Opening Day. Other partnerships announced at the same event — with Accenture, Deloitte, Oracle, and others — also drew attention, but the Merck deal was the biggest story of the week.
Gemini Enterprise had just been redefined as "an integrated platform for AI agents," and this was showcased as its first massive real-world use case.
The contract is a multi-year agreement worth up to $1 billion (~¥150 billion).
Multiple media outlets reported it as "the largest of its kind, both as a single company's AI investment and as a cloud contract in the pharmaceutical industry."
To put it in perspective: for a mid-sized company, this would exceed an entire year's revenue; for Merck, it represents roughly 5% of its annual R&D budget.
In household terms, it's like taking half your annual income and pouring it all at once into appliances and renovations.
Until now, pharmaceutical IT budgets for AI pilot projects typically ranged from tens of millions to a few hundred million yen. This deal pushes the industry benchmark up by two orders of magnitude.
Even Roche's widely noted partnership with NVIDIA was reported at a scale of several hundred million dollars — so a billion-dollar commitment to a single partner rewrites the rules of the industry.
The rollout covers all 75,000 of Merck's employees worldwide.
Gemini Enterprise (Google's enterprise AI platform) will be deployed across "all domains — R&D, manufacturing, commercial (sales and marketing), and corporate (finance, HR, and legal)."
In other words, every researcher, factory manager, and sales rep will have their own dedicated AI assistant. A distinctive feature of the approach is that Google Cloud engineers will be embedded on-site alongside Merck employees to implement AI directly into their workflows.
Rather than a tutor who shows up for three months and leaves, imagine Google's teachers sitting right at your desk, working through problems with you every day. The primary goal is real, on-the-ground AI implementation — not just deployment consulting.
Saying "AI is coming in" doesn't tell us what will actually change. Let's get specific.
The key concept here is agentic AI — AI that operates autonomously. It goes a step beyond the question-answering AI we know from tools like ChatGPT: given a goal, it forms its own plan, operates tools, and delivers results end to end.
If ordinary AI is a receptionist, agentic AI is a new hire.
For example, if you say "Read these 100 papers and create a list of 10 drug candidates," it will search the literature database, read and interpret the papers, and produce a comparison table — all on its own. That's agentic AI.
Gemini Enterprise is the platform (the shared foundation) that enables this to run safely, at scale, and across multiple agents simultaneously.
What's about to happen at Merck is like having an entire squad of AI new hires permanently on staff.
The area generating the most excitement is research and development.
Bringing a single new drug to market typically takes over 10 years: propose candidate molecules → test for efficacy → check for toxicity → animal trials → human clinical trials.
Merck is combining its proprietary AI platform "AIDDISON" with Google Cloud's Vertex AI to run foundation models that predict how molecules behave.
AI pre-screens which molecules are likely to work against a disease and which are likely to cause side effects, narrowing down the field for lab testing. Merck has already demonstrated this kind of impact: in a collaboration with McKinsey, it reduced the time to produce an initial clinical trial report from 180 hours to 80 hours.
It's like playing chess — instead of reading every possible move, AI highlights only the promising lines, so the player can focus on deep thinking.
Beyond R&D, AI will support factory predictive maintenance (repairing equipment before it breaks), sales team customer engagement, legal contract review, and monthly financial reporting.
"In factories, AI predicts when machinery is likely to fail — preventing drug supply disruptions."
"In sales, AI suggests which information about a new drug is most likely to resonate with a particular physician."
"In corporate functions, AI automatically converts quarterly figures into presentation slides."
Think of it like equipping every appliance in your home — the washing machine, refrigerator, and microwave — with smart technology that learns your family's preferences and runs itself accordingly.
The plan is to implement this company-wide: AI handles the background administrative work, while people focus on judgment and communication.
Let's read into the goals through the words of the executives involved.
Merck's Chief Information and Digital Officer, Dave Williams, stated: "AI agents and generative tools will allow teams around the world to reimagine processes at scale and deliver scientific breakthroughs to patients faster."
Speed to patients is the core justification for the investment.
The reality of pharmaceuticals is that for every drug delayed, there may be hundreds of thousands of patients around the world waiting for it.
Merck is also facing the impending patent expiration of its blockbuster cancer drug Keytruda (which generates over ¥3 trillion in annual sales) around 2028, and urgently needs to build its next pillar. This AI investment is more than an efficiency play — it can be read as a strategy for the company's survival.
Google Cloud CEO Thomas Kurian said the goal is to "build the industry's first agentic ecosystem powered by Gemini Enterprise."
This signals ambitions beyond optimizing a single company — it's about creating a blueprint for the entire pharmaceutical industry.
An "agentic ecosystem" means not just one AI assistant per employee, but research, manufacturing, and sales AIs that communicate with each other to get work done.
The difference is between hiring one talented employee versus hiring an entire talented team. Analysts see this as Google's strategy to replicate the model with other major pharma companies — Pfizer, Johnson & Johnson, Novartis — once it proves successful here.
At the same Cloud Next 2026 event, Google announced a new generation of AI chips — "TPU 8t (for training, 3× improvement over the previous generation) / TPU 8i (for inference, 80% performance improvement)" — along with a $750 million (~¥110 billion) support fund for partner companies.
The Merck deal is not simply one company's case study. It's a comprehensive showcase of Google's integrated vision: cloud + chips + AI + hands-on implementation support.
It's like an appliance manufacturer offering not just a new air conditioner, but a full package that includes installation and ductwork. This announcement made clear how seriously Google is pushing back against OpenAI (with Microsoft) and Anthropic (with AWS).
Is Merck running ahead of the pack? Let's look at what others are doing.
Swiss pharmaceutical giant Roche is building what it calls the industry's largest hybrid cloud AI factory in partnership with NVIDIA.
The stated goal is to "improve the efficiency of pharmacovigilance (drug safety monitoring) operations by 50% within 2026."
While Merck is deploying company-wide AI agents through Google, Roche is focused on safety monitoring AI through NVIDIA — different approaches with different emphases.
Major pharmaceutical companies are essentially choosing which AI platform to bet on — like deciding which construction crew to hire when building a house: Google's or NVIDIA's.
Swiss pharma company Novartis has close ties with Anthropic (the developer of Claude). CEO Vas Narasimhan joined Anthropic's board in 2024. Novartis also expanded its partnership with Google's Isomorphic Labs in February 2025, with three drug candidates being jointly explored.
While Merck adopted Google's Gemini platform, Novartis is running a dual strategy with both Anthropic's Claude and Isomorphic Labs.
In browser terms, it's like a power user who runs both Chrome and Safari.
This reflects a broader shift from the era of one company, one AI, toward combining multiple AI systems.
U.S. pharma company Pfizer is partnering with AI research lab Boltz to develop a foundation model for molecular AI — a strategy of owning its own core AI for drug discovery.
Merck is responding on this front too: in December 2025, it released "KERMT" — an open-source small molecule model trained on 11 million molecules — in collaboration with NVIDIA. In February 2026, it announced a collaboration with Mayo Clinic to integrate AI virtual cell technology with clinical and genomic data.
Merck has assembled four cards: AI agents (Google), a drug discovery foundation model (NVIDIA), clinical data AI (Mayo Clinic), and generative AI for drug discovery (Variational AI).
It's as solid as a baseball lineup with pitchers for both sides, a middle reliever, and a closer all locked in.
Can this be dismissed as a story about overseas giants? Let's look at the implications for Japan.
In Japan, Merck operates under the name "MSD K.K. (MSD Japan)," owing to the history of its local partnerships and branding since entering the market in 1955. It's a major pharmaceutical company with approximately 3,000 employees, headquartered near Kitanomaru in Tokyo.
This company-wide AI contract naturally covers MSD Japan as well. Gemini Enterprise is expected to be rolled out in phases to all employees — sales representatives (MRs), researchers, and manufacturing site staff alike.
It's also expected to be used for tasks such as preparing materials for physician education sessions held in Japan and automatically generating drug information documents for healthcare institutions.
This is a classic pattern: business transformation initiated overseas lands at Japanese subsidiaries six months to a year later.
Japan's major pharmaceutical companies are already accelerating their own AI strategies.
Takeda Pharmaceutical has partnered with AWS and NVIDIA to build drug discovery AI. Daiichi Sankyo is using both Google Cloud and Microsoft Azure. Astellas is developing its own proprietary AI platform called Mahol-A-Ba.
However, in terms of concentrating an investment of over ¥100 billion with a single partner, Merck is currently a clear step ahead.
Industry observers say that for Japan's major pharma companies to keep pace at this speed, comparable-scale IT investment within the next few years will be unavoidable.
The question now facing each company is whether they've learned from the lesson of Japanese electronics manufacturers losing global market share by falling behind in the smartphone era.
Japan's healthcare sector has traditionally been cautious about cloud adoption due to the highly sensitive nature of medical data, with AWS and Azure dominating until now. The Merck deal is expected to dramatically raise Google Cloud's profile in the sector.
As a reference case — "Merck spent $1 billion on this" — it becomes one of the most powerful arguments for internal approval processes.
It's the same psychology as feeling more confident taking your family to a restaurant frequented by world-renowned chefs than one recommended by a friend.
From the second half of 2026, there is a strong likelihood that domestic hospital chains, medical data companies, and clinical trial CROs (contract research organizations) will increasingly consider adopting Google Cloud + Gemini.
Yamaguchi-san's days at a university hospital coordinating clinical trials were an endless stream of paperwork.
"Consent forms, reports, follow-ups with physicians, questions from trial participants — even 12-hour days weren't enough."
In the second half of 2026, Gemini Enterprise is introduced to an MSD Japan-led clinical trial.
"Draft responses to LINE messages from trial participants are generated in 10 seconds. Summaries of physician notes are delivered each morning. First drafts of reports are 90% automated."
The result: "The time to trial completion shortened from 18 months to 12 months. I can make it home for dinner with my family five nights a week instead of two."
"It's AI that makes both the patients waiting for new drugs and the staff with families at home happier" — that's how it feels in practice.
Miura-san works at a Tokyo research institute searching for anticancer drug candidates.
"I used to handle everything on my own — reading 1,500 English papers a year, running simulations on 500 molecules, and narrowing it down to 20 candidates."
With Gemini Enterprise and Merck's proprietary AI platform AIDDISON working together, 90% of paper summarization is now automated, and AI pre-scores the viability of each molecule.
"What used to take a month for the first round of screening now takes three days. The time I've freed up goes toward hypothesis generation — the kind of work only a human can do."
"I'm not being worked by the AI. I'm the one working the AI." That's how Miura-san describes it.
This is a prime example of AI making people more creative, not rendering them redundant.
Tanabe-san has lived with diabetes for 15 years.
"Monthly clinic visits, medication explanations, side effect consultations — two hours of waiting was the norm."
After his physician's clinic adopted an AI patient engagement tool provided by MSD Japan, medication questions and side effect checks are handled in advance via a smartphone AI interaction. The in-person appointment is now just a five-minute consultation.
"AI explains new drug releases and whether they're likely to be a good fit for my body in a way I can actually understand."
"I used to think pharmaceutical company AI had nothing to do with me — but both the waiting and the anxiety have decreased."
Reducing the time patients wait for new drugs, and the time they spend in waiting rooms, is the broader social significance of agentic AI.
A. They are part of the same corporate group.
"Merck" is the name used in the United States and Canada; "MSD" is the name used in all other countries, including Japan — but it is the same global pharmaceutical group.
Think of it like Coca-Cola using different brand names in different countries.
Note that Germany's Merck KGaA is a separate, unrelated company (focused on electronic materials and laboratory reagents) — a point of potential confusion. The Google Cloud contract was announced by U.S. Merck (MSD in Japan), and has nothing to do with German Merck KGaA.
A. "Gemini Enterprise is Google's enterprise AI; ChatGPT Enterprise is OpenAI's enterprise AI." The key differences: Gemini can switch between more than 100 AI models and integrates deeply with Google Search, Drive, and Sheets; ChatGPT centers on a single model, with Microsoft 365 integration handled through a separate pathway.
In cooking terms: Gemini is a full course meal; ChatGPT specializes in its signature dish. For organizations like pharmaceutical companies that need to run diverse operations on a single platform, Gemini tends to be the preferred choice.
A. It depends on how you define success, but the industry view is that it's high risk, high reward. A failed drug development program can cost ¥100 billion; a successful one can generate ¥1 trillion in revenue.
Merck's annual R&D budget is approximately $18 billion (~¥2.7 trillion), so this $1 billion commitment represents roughly 5% of that.
Unlike a small factory betting half its annual revenue, for Merck this is just one part of a diversified portfolio. Industry experts calculate that improving the success rate of new drug launches by just 1–2% would be enough to recoup the investment — making it, in their view, a bet that is well worth taking.
A. Clinical data and genetic information originating in Japan must be strictly managed under the Act on the Protection of Personal Information and the revised Next Generation Medical Infrastructure Act.
The practical approach is to store data in Google Cloud's Tokyo and Osaka regions and handle it only in anonymized form.
It would not be permissible, as a general principle, for MSD Japan to use data on Japanese patients to train Gemini through U.S. headquarters.
The era of carelessly sending customer data to overseas servers for training is over. The emerging standard for 2026 is: Japanese data stays in Japan; only the AI is shared globally.
A. For individuals, there is "Gemini for Workspace"; for small and medium-sized businesses, there is "Gemini Business."
Starting at around $22 per person per month (annual contract), these tiers integrate with Gmail, Drive, Docs, and Meet.
The intriguing part: this is essentially the same AI engine that Merck is paying $1 billion for — available to individuals for a few thousand yen a month.
It's like a street-legal SUV sharing its engine bloodline with an F1 race car.
For sole proprietors handling email, small businesses generating quotes, or tutors preparing teaching materials — the same AI working for Merck is now working for you too.
This is not a development that can simply be filed away as "a big overseas contract."
A decision to deploy AI assistants to 75,000 employees with an investment of over ¥100 billion fundamentally changes the rules of competition in the pharmaceutical industry.
This is a clear turning point: from an era of "one new drug, 10 years, ¥300 billion" to an era where "AI lays the groundwork and humans focus on judgment."
Cloud Next 2026 in April 2026 was the moment Google demonstrated its full commitment to AI agents in one of the world's most conservative industries. For Japan's pharmaceutical and healthcare sectors, it is a pivotal moment — one that is urgently pressing each company to decide whether to follow suit or compete on a different axis.
For individual users, the same AI engine Merck is using is already accessible through Gemini Business for a few thousand yen a month.
Start by spending just one day with the Gemini features built into Gmail or Google Docs. That single step is enough to ensure you won't be left behind when the world fully shifts to AI six months from now.
This article is a cross-post from AI Friends.
