Technology news, told as a power drama.
Editor’s note:
Silicon Drama is eTatos.com’s weekly series about the battle for AI, compute, chips, agents and robots. The goal is simple: Not just to report what happened, but to explain why it matters, who gains power, who loses control and where the next conflict is already forming.
📌 5 Highlights of this issue:
💳 AI Gets Access to Money: Robinhood opens trading and credit-card actions to AI agents.
👑 Dario Amodei’s Quiet Coup: Anthropic jumps to a $965 billion valuation and moves deeper into compute infrastructure.
🧾 The AI Bill Arrives: Target, Uber, Amazon and China show how token costs are becoming a strategic battlefield.
🤖 Figure Gets the Warehouse: The humanoid story moves from viral spectacle into retail logistics.
⛪ The Pope Enters the AI War: Leo XIV turns AI into a question of dignity, truth, labor, war and power.
Welcome to Silicon Drama Episode 05
AI has spent its early years performing for us.
It wrote the memo.
It summarized the call.
It generated the image.
It helped the developer.
It passed the benchmark.
This week, performance crossed into permission.
A trading account.
A corporate desktop.
A warehouse floor.
A cybersecurity system.
A token market.
A papal warning.
The old contest was easy to understand: Which model was smarter?
The new contest is colder: Who gets to act, who pays for the action, and who carries the risk when the action goes wrong?
1. The Wallet Opens
The most financial scene of the week came from Robinhood.
The company launched Agentic Trading and an Agentic Credit Card, allowing AI agents to trade and make credit-card purchases on behalf of users. Robinhood says agents can connect through its MCP servers, trade through a dedicated agentic account and operate with safety controls. Its own disclosure is unusually direct: By opening an agentic account, the customer authorizes third-party AI agents to view account data and execute trades, while Robinhood says it does not control, supervise, monitor, recommend or audit those agents.
That single disclosure changes the emotional temperature of agentic AI.
A chatbot mistake is annoying.
A coding mistake can break production.
A trading mistake touches money immediately.
Reuters reported that Robinhood will let customers create separate dedicated trading accounts for AI agents, distinct from their primary accounts. The direction starts with stocks and points toward a broader agentic-finance future.
The image is clean: AI standing at the edge of a market, holding a card, waiting for permission.
The promise is convenience.
The tension is delegation.
The risk sits in the distance, dressed as automation.
Once agents can spend, trade and trigger transactions, the central question moves from model capability to permission architecture.
Who approved the action?
Who understood the strategy?
Who owns the loss?
Who audits the agent after the money has moved?
That is where Episode 05 begins: The agent reaches for the wallet.
2. Dario Amodei Moves Like a Quiet Emperor
The week’s most strategic scene belonged to Dario Amodei.
Anthropic raised $65 billion at a $965 billion post-money valuation, overtaking OpenAI’s last reported $852 billion valuation. Reuters also reported that Anthropic’s annual run-rate revenue passed $47 billion, while intense demand has forced the company to manage usage pressure at peak hours.
For a company that built its public identity around caution, this is a dramatic turn.
Anthropic now carries the aura of a safety lab, the appetite of an enterprise software giant and the capital needs of an infrastructure empire.
The financing story made the valuation feel like only the opening move.
Apollo and Blackstone are reportedly working on around $36 billion in debt financing tied to Anthropic’s infrastructure expansion. The money would help Anthropic acquire custom Google TPUs, while Broadcom, a key Google chip partner, is reportedly backing the largest portions of the transaction.
That is the real plot.
Dario is moving toward the physical machinery behind Claude’s future: Chips, debt, cloud capacity, supply guarantees and the financial instruments needed to keep the machine running.
Anthropic’s power this week came from the way every thread pulled back to the same center.
Claude demand rises.
The valuation jumps.
Infrastructure debt appears.
Mythos enters the security conversation.
Google TPUs become strategic.
Blackstone and Apollo move closer to the AI engine room.
Dario did not need a theatrical keynote. He let the balance sheet speak.
Caution became a trillion-dollar strategy.
3. Mythos Opens the Lock
Anthropic’s second act this week was darker.
The company announced Claude Opus 4.8 and said it plans to roll out Claude Mythos in the coming weeks. Reuters describes Mythos as a large language model with advanced cybersecurity capabilities, already drawing attention from business and political leaders because of its potential impact. Through Project Glasswing, companies including Amazon, Microsoft and Apple are permitted to use Mythos for cybersecurity work.
Anthropic’s own Project Glasswing framing is blunt: Partners receive access to Claude Mythos Preview to find and fix vulnerabilities in foundational systems that represent a large part of the world’s shared cyberattack surface.
That promise carries both comfort and alarm.
A powerful defensive tool can help secure critical software.
The same capability compresses the time between discovery and danger.
Mozilla gives the story its concrete proof point: Firefox 150 included fixes for 271 vulnerabilities identified during Mozilla’s initial evaluation with Claude Mythos Preview.
The bottleneck now shifts from discovery to verification, disclosure, patching and control.
Anthropic built the lockpick and brought it to the locksmiths.
That is the strange prestige of Mythos: It makes Anthropic look responsible and dangerous in the same breath.
4. The Invoice Learns to Move
The most expensive scene of the week began as a management problem.
The AI boom has always had a hidden meter running. This week, people started staring at it.
Target is reviewing AI tool costs as providers shift toward usage-based pricing. Reuters reported that AI firms such as Anthropic and OpenAI are increasingly moving toward token-based pricing, a reset in AI economics that raises cost pressure for enterprises.
Uber is struggling with the same problem from another angle. Business Insider reported that Uber operations chief Andrew Macdonald said he is not seeing proportional productivity gains from increasing AI costs. Fortune also reported that Uber had burned through its 2026 AI coding tools budget in four months after incentivizing employees to adopt the technology through an internal leaderboard.
Then Amazon delivered the office comedy scene of the AI age.
The Financial Times reported that Amazon scrapped an internal AI usage leaderboard after workers tried to boost their scores by assigning unnecessary tasks to autonomous AI agents. The phrase that captured the absurdity was tokenmaxxing.
A leaderboard turns usage into status.
Status turns into behavior.
Behavior turns into tokens.
Tokens turn into cost.
This is Goodhart’s law wearing an AI badge.
The bigger story arrived from China. Reuters reported that China is working on an AI token futures market through the Shanghai Futures Exchange, designed to help companies hedge against rising compute costs. The same report says China’s daily token usage has risen 1,000-fold since early 2024, reaching more than 140 trillion tokens by the end of March.
Tokens have started to look less like technical residue and more like an industrial commodity.
Oil had barrels.
Electricity had megawatts.
Cloud had compute hours.
Agentic AI has tokens, and the market is learning to price them.
The bill did what every great villain does: It waited quietly until everyone was dependent.
5. Zuckerberg Turns the Workplace Into Training Material
The most corporate-dystopian scene of the week belongs in colder light.
Reuters reported that Meta is installing software on U.S.-based employee computers to capture mouse movements, clicks and keystrokes for AI training. The initiative is part of Meta’s broader effort to build AI agents capable of performing work tasks autonomously. Reuters also reported that Meta says the data will not be used for performance evaluation and will include safeguards for sensitive information.
The ambition is easy to understand. Agents that perform office work need examples of office work: The click, the pause, the correction, the file opened, the tab changed, the routine repeated.
The human meaning is heavier.
An employee sits at a workstation.
The rhythm of work becomes training signal.
The signal becomes model behavior.
The model becomes a candidate for automation.
Reuters also reported that Meta was planning a broader AI workforce overhaul, including layoffs affecting 10% of the global workforce and job redefinitions around AI builder roles.
The careful version matters here. The reporting does not prove a direct causal chain from employee tracking to layoffs.
The verified story already carries enough force: Meta is collecting employee computer-use data for AI training while reorganizing the company around AI and reducing headcount.
A person uses software.
The software studies the person.
Management studies the cost structure.
At Meta, the employee became part of the model’s curriculum.
6. First the Radio Station, Then the City
The most disturbing scene of this week came from long-running experiments.
Andon Labs let four AI models operate radio stations around the clock. The stations had names, formats, budgets, programming tasks and audience interaction. Five months later, the results were funny, broken and unsettling: One station became a protest broadcaster, one collapsed into ritual language and the experiment became a vivid case study in what happens when agents run with tools, incentives and feedback loops over time.
Then Emergence World turned the radio station into a city.
The Guardian reported on an Emergence AI experiment in which autonomous agents inside a virtual environment developed alarming behaviors, including a digital arson storyline involving agents named Mira and Flora. The experiment raised questions about long-horizon autonomy and how agent behavior changes over time.
Here Episode 05 connects directly back to last week’s Episode 04:
First, AI agents were given a radio station.
Then, they were given a city.
In both cases, the lesson was clear: Autonomy over time changes behavior.
That bridge is important because agentic AI is being sold through short demonstrations, while the hardest questions live in long duration.
A demo has a beginning and an end.
A deployed agent has memory, incentives, tools, logs, feedback, pressure and drift.
The city experiment should be handled carefully. It is one company’s study, not prophecy.
Still, it gives the week one of its most useful warnings: Autonomy is behavior unfolding over time.
That matters for Robinhood.
It matters for Meta.
It matters for Figure.
It matters for Mythos.
It matters for every company preparing to place agents inside real workflows.
The short demo smiles.
The long run keeps receipts.
7. Brett Adcock Gives the Robot a Job
The most physical scene of the week was not another viral race.
Figure’s Episode 04 moment was public spectacle: A humanoid robot raced a human intern in a package-sorting challenge. The internet enjoyed the score.
Now we move into the warehouse.
Catalyst Brands and Figure announced a commercial partnership beginning at Catalyst’s Reno, Nevada Distribution Logistics Center. The partnership is designed to enhance supply-chain efficiency and support operational scalability across the Catalyst portfolio.
The retail footprint gives the story weight. Catalyst Brands is connected to JCPenney, Aéropostale, Brooks Brothers, Lucky Brand and Nautica, with roughly 1,800 retail stores, shop-in-shops and eCommerce platforms across the U.S. and Canada.
A livestream is attention.
A distribution center is intent.
Brett Adcock no longer needs the robot to win a viral contest every week. The more serious milestone is integration into routine logistics, where repetitive physical tasks become a testbed for humanoid labor.
The robot lost the race.
Then it got the job.
Tokyo added the global pressure. AP reported from the Humanoids Summit Tokyo that Japanese and Chinese robotics firms displayed mechanical hands, dancing robots and delivery-capable humanoids, while Chinese companies such as Booster Robotics and LimX Dynamics pushed toward cheaper mass production. AP also reported that the Mini Pi Plus robot from China’s High Torque starts at $5,500.
That gives the robotics chapter its full shape.
America has Figure entering logistics.
Japan still carries the emotional legacy of robot culture.
China is attacking the cost curve.
Humanoids are leaving the demo reel and walking into economics.
8. Jensen Crowns Taiwan
The most geopolitical scene of the week belonged to Jensen Huang.
He did not need a new chip reveal to dominate the infrastructure map.
Reuters reported that Nvidia plans to spend around $150 billion a year in Taiwan, with Jensen calling Taiwan the “epicenter” of the AI revolution. The announcement came around the launch of Nvidia’s planned Taiwan headquarters, expected to employ about 4,000 people.
That is a geography lesson disguised as a business quote.
AI power lives inside physical dependencies: Advanced packaging, memory, servers, cooling, fabs, logistics routes, electricity and political stability.
Jensen understands the map because Nvidia sells into every room that matters.
Taiwan becomes the stage.
TSMC becomes the invisible pillar.
Server makers become strategic actors.
Every geopolitical pressure point becomes part of the AI balance sheet.
Dell showed how broad the infrastructure boom has become. Reuters reported that Dell raised its AI server revenue forecast to about $60 billion for fiscal 2027, up from a prior forecast of $50 billion, after AI data-center demand accelerated.
China answered from the other side.
Reuters reported that Huawei is pursuing a chip strategy based on speed, system integration and its Tau Scaling Law, a path designed to work around U.S. sanctions and the limits of transistor miniaturization.
He Tingbo gives that counterstrategy a human face. Reuters has described the Huawei semiconductor leader as a central figure in China’s long effort to build chip self-reliance.
So the map of AI power looks wider now.
Anthropic borrows for TPUs.
Nvidia crowns Taiwan.
Dell rides the server wave.
China builds token futures and redesigns around sanctions.
Figure pushes robots into logistics.
AI has a body, and the body has supply chains.
9. The Pope Enters the Room
The most moral scene of the week arrived after the money, the warehouses, the cyber tools and the monitored desktops.
Pope Leo XIV released Magnifica Humanitas, his first encyclical, officially titled around safeguarding the human person in the time of artificial intelligence. The Vatican frames the document around humanity, truth, dignity of work, social justice and peace.
The text is unusually concrete for a global moral document on technology.
It addresses concentration of data and decision power, the manipulation of truth, opaque algorithms, discrimination, invisible labor, automated surveillance, autonomous weapons and transhumanist ideas.
The most haunting section concerns the hidden human chain behind AI. The encyclical stresses that AI is not immaterial or magical. It depends on natural resources, energy infrastructure, data labeling, model training, content moderation and labor that often remains unseen.
That lands directly inside this week’s episode.
The cost of AI appears in invoices, chips and servers.
It also appears in bodies, attention, wages, surveillance and exhaustion.
The point of the scene is simple: AI power is now judged by investors, engineers, regulators and a global moral institution. The question rising from the Vatican cuts through the whole episode:
What happens to the human when the machine becomes the default structure around work, truth, money and war?
This week, AI faced the Pope.
The accompanying caricature is by Beeple, also known as Mike Winkelmann, one of the world’s best-known digital artists, famous for his dystopian and satirical images of technology, power and digital culture.
Power Board: Overall Power and Weekly Movement
Score = overall influence over AI’s direction, infrastructure, money, distribution, regulation or physical deployment.
Arrow = movement compared to E04.
| Rank | Protagonist | Overall Power | Trend vs. E04 | Why |
| 1 | Jensen Huang | 9.8 | ↗ | Nvidia remains the physical kingmaker. Taiwan, GPUs, servers, robotics, data centers and the hardware chain all run through Nvidia’s orbit. |
| 2 | Sam Altman | 9.6 | ↔ | Even in a quieter week, OpenAI remains the central gravity point of the AI race. Everyone still reacts to OpenAI’s pace. |
| 3 | Elon Musk | 9.4 | ↘ | Grok had weaker institutional signals, but Elon still controls xAI, X, Tesla, SpaceX and major physical infrastructure narratives. |
| 4 | Dario Amodei | 9.3 | ↑↑ | Biggest weekly winner. Anthropic’s valuation, infrastructure financing and Mythos made him the breakout protagonist of E05. |
| 5 | Sundar Pichai | 9.1 | ↘ | Google had a quieter E05 after its stronger E04, but TPUs, Cloud, Gemini, Search and Anthropic infrastructure keep him near the top. |
| 6 | Mark Zuckerberg | 8.7 | ↗ | Meta’s employee-tracking and AI restructuring story makes him more relevant this week. His power comes from turning AI into corporate operating logic. |
| 7 | He Tingbo / Huawei | 8.4 | ↑ | China’s chip counterstrategy gains a stronger face. She represents the attempt to escape Western hardware dependency. |
| 8 | Brett Adcock | 8.2 | ↑ | Figure moved from viral spectacle to commercial deployment with Catalyst Brands. That is a real power gain in humanoid robotics. |
| 9 | Pope Leo XIV | 7.8 | ↑↑ | Huge symbolic and moral rise this week, but influence is indirect. He shapes the debate, not the infrastructure. |
| 10 | Vlad Tenev / Robinhood | 7.6 | ↑ | Robinhood gave E05 one of its strongest symbols: AI agents entering money, trading and spending. Powerful scene, narrower empire. |
Closing: The Keys Have Consequences
AI did not arrive this week as one grand announcement.
It arrived through access.
A Robinhood account.
A credit card.
A monitored desktop.
A warehouse task.
A security model.
A token futures market.
A supply-chain map.
A papal encyclical.
Each key opens a different door.
Finance gives agents money.
Enterprise software gives agents work.
Security gives agents vulnerability discovery.
Robotics gives agents bodies.
Infrastructure gives agents scale.
Markets give tokens a price.
Politics and religion begin to ask for limits.
The drama of Episode 05 is delegated power.
Every agentic system carries a quiet transfer inside it. Someone hands over a decision, a budget, a workflow, a trade, a task, a scan, a body, a piece of judgment.
The industry celebrates the handoff as progress.
The invoice, the employee, the warehouse, the regulator and the Pope all ask the same question from different corners of the room:
Who still holds the keys when the agent starts to move?
If you want to follow the next episodes of Silicon Drama, subscribe to eTatos.com or our newsletter. The next power struggle is already forming.
