The Day One CEO Shook Nvidia's Share Price── A CEO's apology, a deal to stop AI, pharma AI's confirmed sums
Download the video (MP4, 11 MB)
News about AI usually arrives with a big sum attached. But a headline alone rarely tells you how solid that sum is. Today's events showed what AI's worth really rests on, seen from three places: investors, governments, and the drug industry. Drug companies are also budgeting for AI and slowly handing it work. So this question concerns every department, not only research. How do we judge by substance, without being pulled in by big numbers?
Today I start with the turmoil among investors and follow the answer down to everyday work in our own industry.
01Nvidia's CEO Apologizes
Source AP News ／ Forbes ／ Fortune
When investors judge an AI company, they look first at the price put on it. But what holds that price up? Today's reports suggested that much of it rests on trust in a single person.
What struck me here is the gap between a small trigger and a huge shock. Amid turmoil around the person at the top, that leader said sorry, investors went to court, and the company's value was cut sharply. As reported, what shook was not the technology. It was trust in that leader.
The scale of what vanished tells me investors had placed much of the company's worth on one individual. And what actually happened is still not clear from the headlines. The price moved while the substance stayed unknown.
The same risk applies when a drug maker picks an AI tool. If a developer's public face is one person, a scandal around them could reach your contract. If its stock swings, development pace and pricing may follow. So alongside a tool's performance, I would check whether its maker leans too hard on one leader.
If trust in one leader is that fragile, what else gives a company that builds AI its real worth?
02Claude Inside a Rival's Screen
Source The Verge ／ The Register ／ Dealroom
When a company's worth leans on one leader, it can shake this hard. So what else does an AI maker's worth lean on? Today's moves pointed to one thing: whether it can get inside the tools people open every day.
Its own office suite
The Verge reported Anthropic has launched its own AI-powered office suite. A move that could compete directly with Microsoft.
Claude inside Copilot
The Register reported that Microsoft will enable Claude by default in UK Copilot tenants.
Offers relayed by AI
Dealroom reported Ramp hid an AI-only offer on 50 web pages. Claude relayed it 370 times in 32 days; ChatGPT, 0 times.
What these moves share is that makers are going after the entry point in users' hands. One maker fights head on with its own tools, yet also chooses to work within a competitor's. As reported, competing and partnering are happening at once. AI has also begun to pick up information and pass it on to people. I read this as the place where things are sold shifting toward AI.
A tool that works without the user choosing anything holds real power. Few people bother to switch. Being there from the start likely shapes who earns more.
In drug companies too, AI will increasingly arrive built into everyday office software. Then which AI receives internal documents may be decided without anyone choosing. Teams handling patient data or unpublished material should confirm which AI is running before they start.
The deeper AI sits in daily tools, the more work it touches. That raises the question of whether it can be halted when needed.
03Mechanisms to Stop AI
Source The Decoder ／ 24/7 Wall St. ／ Middle East Monitor
As AI grows stronger, the question shifts from what it can do to whether it can be halted in an emergency. Today, governments and developers moved closer on that point.
This sequence shows safety moving from promises on paper to a working device. Around the time governments signed on, one developer was said to have created a way to halt its own systems quickly. Even so, some argue that an accord alone has little force, and they want rules with teeth. I see the ability to halt becoming a condition for selling.
The time it takes was reported as under a minute. But in what situations, and on whose call, is still not visible from the headlines. Speed alone cannot tell us how safe we really are.
In pharma, drafts and decision suggestions from AI are entering work that affects patients. When it misbehaves, who halts it, how, and what gets recorded? Deciding that before rollout is a safeguard that does not rely on the maker's promise. And the procedure should be rehearsed once, not just written down.
Even with that safeguard in place, the next question is the money needed to keep using that AI.
04Cost per Task over Performance
Source Google DeepMind ／ OrcaRouter ／ OrcaRouter
Each new model brings talk of how much smarter it is. But what buyers really want to know is whether that intelligence is worth the bill.
| Compared with | Gap in performance / claims | Cost per task |
|---|---|---|
| DeepSeek V4 Pro | 37 index points higher | 4.4x more |
| Grok 4.7 | 3x on paper | About equal per task |
Third-party comparison by OrcaRouter. The axis of comparison is shifting from raw performance to cost per task.
This comparison shows that a lead in intelligence does not turn straight into value. Against one rival, a big win comes with a bill several times larger to get the job done. Against another, a wide gap in headline numbers shrinks to roughly the same spend for each job. Intelligence only means something when set beside what you pay.
A single query may be cheap. But if you must ask again and again to reach an answer, the bill climbs. How many retries you need depends on how hard the work is. So pricing a whole finished job is closer to reality.
Drug companies want to hand AI many jobs, such as literature summaries and draft regulatory filings. Choosing a model by ranking alone can make spending balloon. I would trial our own jobs one by one and compare the spend to finish each. Using top models only for demanding work is another option.
Looking at spend job by job also helps when reading how much AI money is flowing into our industry.
05Pharma AI's Confirmed Sums
Source Pulse 2.0 ／ Pulse 2.0 ／ Business Wire
Talk of drug makers and AI comes with big figures too. But how much of that money has actually been paid in?
| Deal | Amount | Nature of amount |
|---|---|---|
| Series A of an AI drug discovery firm | $55 million | Confirmed raise |
| Series C of an AI drug discovery firm | ~$15 million (RMB 100 million) | Confirmed raise |
| Proximal Ventures' debut fund | $27.5 million | Plan / target, not a confirmed sum |
| Big pharma–SAP partnership expansion | Amount not disclosed | Scale cannot be judged |
The money went not to big pharma's own investment but to private firms with AI drug discovery platforms.
What matters here is less the size of each figure than what kind of figure it is. Money already committed and money someone hopes to raise weigh differently, even side by side. A tie-up with no figure cannot be called big or small. Adding up headline numbers distorts the scale. Money that moved and money planned to move must be read apart.
Both firm sums went to young companies that have not listed their shares. Rather than big drug makers betting their own funds, outside investors backed firms with systems for finding drug candidates. In today's material, the big players showed up in operational tie-ups with no price attached.
When drug makers partner with such firms, the partner is still raising money. If the next round fails, joint research may stall. Checking where their money comes from, and how long it will last, protects your own plans.
Apart from money, AI was also entering drug companies in other ways, outside the lab.
06On the Pharma Floor
Source SAP News Center ／ EIN News ／ MONOist
Hear AI and drug makers together, and most people picture the hunt for new drug candidates. Today's moves happened somewhere else.
Business backbone
A big pharma company expands its SAP partnership to adopt RISE with SAP and Business AI. AI goes into operations, not drug discovery.
Microbiology testing
PSC Biotech and Clever Culture Systems partner to bring AI-enabled microbiology test automation to pharma manufacturers.
Cell culture media
Shimadzu and US-based Axio have begun a collaboration to optimize cell culture media using AI.
None of these moves is about discovering medicines. They cover the systems that run daily operations, the checks that confirm quality, and the mix of ingredients used in production. Together they show AI leaving the lab for routine work and manufacturing. Perhaps results are easier to prove quickly in daily operations than in research.
Setting the recipe for growing cells takes a lot of effort. With AI, that effort may shrink. But how a changed recipe affects product quality must keep being checked by people.
Once AI enters manufacturing and quality testing, the topic moves closer to regulation than research. Test results decide whether a product may be released. Keeping the reasoning of AI explainable after the fact will become new work for quality teams. And AI used in testing must be validated, just like the test equipment itself.
The deeper AI goes into company work, the more a different question arises for the people who work there.
07AI-Driven Cuts and a Court Ruling
Source NNAアジア ／ Quartz India ／ UA.NEWS
If AI takes on work, what happens to people? Today, companies and a judge gave different answers.
| Who | Move | What it means |
|---|---|---|
| Qantas | Cutting about 850 jobs in a deal with Accenture | Outsourcing IT work and expanding AI use |
| Axonius | Cut 22% of its developers in Israel | Reported as an AI-driven restructuring |
| Taiwan district court | Ruled AI-based firing unfair; ordered Microsoft Taiwan to pay damages | A court rejecting the stated reason for cuts |
| Andrea Nahles | Predicts 800,000 job losses in Germany due to AI | A forecast, not actual cuts |
Companies are cutting staff in the name of AI, and courts have begun to rule some of it unfair.
On the corporate side, staff reductions justified by AI keep coming. Some hand work to outside firms. Others shrink engineering teams in one stroke. Amid this, one judge refused to accept AI as grounds for letting someone go, and told the employer to compensate. At least in this case, the label of AI alone was not accepted as a reason. Acting on company convenience alone can be challenged in law.
Drug companies will also redesign work around AI. Then the reason for moving people must rest on how the work itself changed, not on AI. Unexplained changes lose employees' trust and may end up before a judge. Talking with each person early about how their work will change helps prevent disputes, I believe.
Deciding to let people go is not about how capable AI is. It is about company responsibility. What you told employees, and how, will be examined later.
The real worth of AI only comes into view when you include how a company treats its people.
Watch whether the stop agreement in the Paris declaration moves into actual operation.
Open the full transcript
Intro
News about AI usually arrives with a big sum attached. But a headline alone rarely tells you how solid that sum is. Today's events showed what AI's worth really rests on, seen from three places: investors, governments, and the drug industry. Drug companies are also budgeting for AI and slowly handing it work. So this question concerns every department, not only research. How do we judge by substance, without being pulled in by big numbers?
Today I start with the turmoil among investors and follow the answer down to everyday work in our own industry.
CH 01 Nvidia's CEO Apologizes
When investors judge an AI company, they look first at the price put on it. But what holds that price up?Today's reports suggested that much of it rests on trust in a single person.What struck me here is the gap between a small trigger and a huge shock. Amid turmoil around the person at the top, that leader said sorry, investors went to court, and the company's value was cut sharply. As reported, what shook was not the technology. It was trust in that leader.The scale of what vanished tells me investors had placed much of the company's worth on one individual. And what actually happened is still not clear from the headlines. The price moved while the substance stayed unknown.The same risk applies when a drug maker picks an AI tool. If a developer's public face is one person, a scandal around them could reach your contract. If its stock swings, development pace and pricing may follow. So alongside a tool's performance, I would check whether its maker leans too hard on one leader. If trust in one leader is that fragile, what else gives a company that builds AI its real worth?
CH 02 Claude Inside a Rival's Screen
When a company's worth leans on one leader, it can shake this hard. So what else does an AI maker's worth lean on?Today's moves pointed to one thing: whether it can get inside the tools people open every day.What these moves share is that makers are going after the entry point in users' hands. One maker fights head on with its own tools, yet also chooses to work within a competitor's. As reported, competing and partnering are happening at once. AI has also begun to pick up information and pass it on to people. I read this as the place where things are sold shifting toward AI.A tool that works without the user choosing anything holds real power. Few people bother to switch. Being there from the start likely shapes who earns more.In drug companies too, AI will increasingly arrive built into everyday office software. Then which AI receives internal documents may be decided without anyone choosing. Teams handling patient data or unpublished material should confirm which AI is running before they start. The deeper AI sits in daily tools, the more work it touches. That raises the question of whether it can be halted when needed.
CH 03 Mechanisms to Stop AI
As AI grows stronger, the question shifts from what it can do to whether it can be halted in an emergency. Today, governments and developers moved closer on that point.This sequence shows safety moving from promises on paper to a working device. Around the time governments signed on, one developer was said to have created a way to halt its own systems quickly. Even so, some argue that an accord alone has little force, and they want rules with teeth. I see the ability to halt becoming a condition for selling.The time it takes was reported as under a minute. But in what situations, and on whose call, is still not visible from the headlines. Speed alone cannot tell us how safe we really are.In pharma, drafts and decision suggestions from AI are entering work that affects patients. When it misbehaves, who halts it, how, and what gets recorded?
Deciding that before rollout is a safeguard that does not rely on the maker's promise. And the procedure should be rehearsed once, not just written down. Even with that safeguard in place, the next question is the money needed to keep using that AI.
CH 04 Cost per Task over Performance
Each new model brings talk of how much smarter it is. But what buyers really want to know is whether that intelligence is worth the bill.This comparison shows that a lead in intelligence does not turn straight into value. Against one rival, a big win comes with a bill several times larger to get the job done. Against another, a wide gap in headline numbers shrinks to roughly the same spend for each job. Intelligence only means something when set beside what you pay.A single query may be cheap. But if you must ask again and again to reach an answer, the bill climbs. How many retries you need depends on how hard the work is. So pricing a whole finished job is closer to reality.Drug companies want to hand AI many jobs, such as literature summaries and draft regulatory filings. Choosing a model by ranking alone can make spending balloon. I would trial our own jobs one by one and compare the spend to finish each. Using top models only for demanding work is another option. Looking at spend job by job also helps when reading how much AI money is flowing into our industry.
CH 05 Pharma AI's Confirmed Sums
Talk of drug makers and AI comes with big figures too. But how much of that money has actually been paid in?What matters here is less the size of each figure than what kind of figure it is. Money already committed and money someone hopes to raise weigh differently, even side by side. A tie-up with no figure cannot be called big or small. Adding up headline numbers distorts the scale. Money that moved and money planned to move must be read apart.Both firm sums went to young companies that have not listed their shares. Rather than big drug makers betting their own funds, outside investors backed firms with systems for finding drug candidates. In today's material, the big players showed up in operational tie-ups with no price attached.When drug makers partner with such firms, the partner is still raising money. If the next round fails, joint research may stall. Checking where their money comes from, and how long it will last, protects your own plans. Apart from money, AI was also entering drug companies in other ways, outside the lab.
CH 06 On the Pharma Floor
Hear AI and drug makers together, and most people picture the hunt for new drug candidates. Today's moves happened somewhere else.None of these moves is about discovering medicines. They cover the systems that run daily operations, the checks that confirm quality, and the mix of ingredients used in production. Together they show AI leaving the lab for routine work and manufacturing. Perhaps results are easier to prove quickly in daily operations than in research.Setting the recipe for growing cells takes a lot of effort. With AI, that effort may shrink. But how a changed recipe affects product quality must keep being checked by people.Once AI enters manufacturing and quality testing, the topic moves closer to regulation than research. Test results decide whether a product may be released. Keeping the reasoning of AI explainable after the fact will become new work for quality teams. And AI used in testing must be validated, just like the test equipment itself. The deeper AI goes into company work, the more a different question arises for the people who work there.
CH 07 AI-Driven Cuts and a Court Ruling
If AI takes on work, what happens to people?Today, companies and a judge gave different answers.On the corporate side, staff reductions justified by AI keep coming. Some hand work to outside firms. Others shrink engineering teams in one stroke. Amid this, one judge refused to accept AI as grounds for letting someone go, and told the employer to compensate. At least in this case, the label of AI alone was not accepted as a reason. Acting on company convenience alone can be challenged in law.Drug companies will also redesign work around AI. Then the reason for moving people must rest on how the work itself changed, not on AI. Unexplained changes lose employees' trust and may end up before a judge. Talking with each person early about how their work will change helps prevent disputes, I believe.Deciding to let people go is not about how capable AI is. It is about company responsibility. What you told employees, and how, will be examined later. The real worth of AI only comes into view when you include how a company treats its people.
Wrap-up
Every event today asked what is solid behind the big numbers. Intelligence and deal size are not enough to judge AI. Whom does it depend on, where does it enter, can it be halted, what does each job cost, and who answers for it?
These are the questions to ask when choosing AI in our work. Tomorrow, we see whether the accord between governments and developers gets used for real.
- CH 01AP News「Nvidia and Jensen Huang face shareholder lawsuit alleging investors were misled about spy affair」 apnews.com
- CH 01Forbes「Nvidia CEO Jensen Huang’s Apology Provides Crisis Communication Lessons」 forbes.com
- CH 01Fortune「Jensen Huang is the AI boom’s key man. Now, that’s a problem」 fortune.com
- CH 02The Verge「Anthropic has its own AI-powered office suite now」 theverge.com
- CH 02The Register「Microsoft is giving Claude the keys to UK Copilot tenants by default」 theregister.com
- CH 02Dealroom「Ramp hid an AI-only offer on 50 web pages: Claude relayed it 370 times in 32 days, ChatGPT never did」 app.dealroom.co
- CH 03The Decoder「Eight countries and top AI labs agree to shutdown mechanisms in new Paris declaration」 the-decoder.com
- CH 0324/7 Wall St.「OpenAI Built a Kill Switch That Can Stop Its AI Models Within 35 Seconds」 247wallst.com
- CH 03Middle East Monitor「Don’t wait for the blackout: The world needs binding rules for dangerous AI」 middleeastmonitor.com
- CH 04Google DeepMind「Gemini 4 Carbon is now available」 deepmind.google
- CH 04OrcaRouter「Gemini 4 Carbon vs DeepSeek V4 Pro: What 37 Index Points and 4.4x the Cost Per Task Actually Buy」 orcarouter.ai
- CH 04OrcaRouter「Gemini 4 Carbon vs Grok 4.7: 3x on paper, parity per task」 orcarouter.ai
- CH 05Pulse 2.0「Fathom Therapeutics Raises $55 Million Series A Led By Sutter Hill Ventures To Advance AI Drug Discovery」 pulse2.com
- CH 05Pulse 2.0「Sanyou Bio: Nearly $15 Million (RMB 100 Million) Series C Funding Expands AI-Powered Drug Discovery Platform」 pulse2.com
- CH 05Business Wire「Proximal Ventures Announces $27.5 Million Debut Fund to Invest in AI-Native Companies Reshaping Life Sciences」 businesswire.com
- CH 06SAP News Center「SAP Expands Collaboration with AstraZeneca to Adopt RISE with SAP and Business AI」 news.sap.com
- CH 06EIN News「PSC Biotech Partners with Clever Culture Systems to Bring AI-Enabled Microbiology Automation to Pharma Manufacturers」 einnews.com
- CH 06MONOist「島津製作所、米AxioとAIを活用した細胞培養培地最適化の協業を開始：医療技術ニュース」 monoist.itmedia.co.jp
- CH 07NNAアジア「ＡＩ理由の解雇は不当、台湾ＭＳに地裁賠償命令」 nna.jp
- CH 07Quartz India「Qantas is cutting about 850 jobs in an AI-driven deal with Accenture」 qz.com
- CH 07UA.NEWS「Andrea Nahles predicts loss of 800,000 jobs due to AI — Der Spiegel」 ua.news
Articles used
Today's related reports
- AI Daily News November 11, 2026
- AI & Economy News November 11, 2026
- AI & Finance News November 11, 2026
The narration is synthetic speech. The content draws only on this site's articles from the same day. Each edition passes seven plain-language checks before publication. That means known obstacles to comprehension are held below threshold — it is not a guarantee of comprehension.