A single AI company is systematically

dismantling the entire software industry

>> [music]

>> and they're doing it on a predictable

schedule. In January, Claude's legal

plug-in wiped out $300

worth of legal software stocks. In

February, Opus 4.6 triggered the

trillion-dollar SaaS apocalypse. In

March, Anthropic said that Mythos was

too dangerous to release and

cybersecurity stocks crashed anyway. And

now, Anthropic just dropped Claude

design and dropped stocks like Adobe,

Figma, and Wix along with it. Four

months, four major industries, and it's

only speeding up from here. My name is

Alex and I spent eight years as an

electrical engineer and AI researcher at

MIT and I've never seen AI move this

fast. So, subscribe to the channel and

let me show you the bigger pattern and

how I'm investing in it. Your time is

valuable, so let's get right into it.

The strange thing about Claude design is

that it doesn't look like the end of all

software. It just looks like another

design tool. You open Claude, you

describe what you want, like a landing

page, a dashboard, a pitch deck, or a

prototype, and a few seconds later, the

model gets you something usable. Not a

chat response and not a list of

suggestions, an artifact that you can

edit, export, or hand off to Claude

code, which is another agentic AI tool

that recently rattled software stocks.

If you're a professional designer, you

probably think that Claude design is

cute. It has generic layouts, bland

taste, and poor design. So, it's not

exactly replacing senior designers at

Adobe or Apple anytime soon. That's the

right reaction and it also completely

misses the point. Figma stock didn't

drop because Claude design is already

better than Figma. Adobe didn't fall

because Anthropic replaced Photoshop and

Wix didn't get hit because Claude is a

better website builder. The market

reacted because Claude design is part of

a bigger pattern. First, legal software,

then enterprise software, then security

software, and now design software. Four

high-margin software industries in the

last 90 days. One company, one clear

market pattern. On January 12th,

Anthropic launched Claude Co-work, which

is basically Claude code, but for the

rest of your work. This isn't just a

coding assistant, it's a model inside

your normal workflows. It can read and

edit files, it can run inside a sandbox,

and it can produce work instead of just

talking about it. Then, on January 30th,

Anthropic expanded Co-work with 11

open-source plugins across huge markets

like HR, design, engineering,

operations, financial analysis,

investment banking, stock market

research, wealth management, legal, and

of course, marketing. That sounds like a

product roadmap, but for investors, it's

something much more dangerous. Each of

those software categories is full of

public companies that built their entire

business around humans clicking through

workflows one seat at a time. By early

February, the market had a name for what

happened next, the SaaS apocalypse.

Around $300 billion of software market

value got wiped out in a matter of weeks

and some haven't yet recovered. That

wasn't the entire software industry

dying, it was the stock market asking an

important question. If Claude can do the

work inside my software, then why do

companies need so many expensive seats?

And that was only the first strike. On

February 20th, Anthropic announced

Claude code security. Now, the target

was finding bugs and exploits in code.

The model could scan codebases, identify

vulnerabilities, and even propose

patches for humans to review.

Cybersecurity stocks sold off because

investors understood the implications

immediately. If an AI agent can go

straight into the codebase and do

security work directly, that changes the

value of cybersecurity companies and

expensive consultants. Then, the pattern

escalated. On March 26th and 27th,

Claude Mythos was leaked, a model

Anthropic knew was far ahead of anything

else in terms of its cyber capabilities.

The leak was an unprecedented

cybersecurity risk, so cybersecurity

stocks like CrowdStrike, Palo Alto

Networks, and Zscaler all sold off. The

model wasn't even released to the public

because it was so dangerous and it moved

the market anyway. Then, on April 7th,

Anthropic created Project Glasswing.

Mythos was given to companies like AWS,

Apple, Broadcom, Cisco, and CrowdStrike,

Google, Microsoft, Nvidia, and Palo Alto

Networks. I covered this in my most

recent video, but it's too important not

to tell you again. Mythos found

thousands of huge software exploits,

including one 27-year-old bug that could

crash any machine just by connecting to

it. Mythos is not a prototype. Mythos is

a weapon and that brings us to today.

Let's talk about Claude design, which

just dropped a few days ago. When stocks

like Adobe and Figma drop, it's not the

market responding to a single product,

it's responding to a timeline. January

12th, Claude Co-work. February 20th,

Claude code security. March 26th,

Mythos. April 17th, Claude design. By

the time investors finish arguing about

whether any of these launches are

overhyped, the next launch is already

hitting a different software category.

And that's where the clock begins. From

April 18th to the end of 2026, there are

257

days. Anthropic has been shipping major

product moves roughly every two weeks

and there are about 26 major software

categories across the entire economy.

Things like project management,

procurement, and marketing, all the way

to sales, customer support, finance, and

coding. And it's not just about

Anthropic. If any frontier AI labs keep

hitting software categories like this

every two weeks, the market should see

around 18 more shocks like this before

the end of the year. And every single

one of them asks the same question, is

this software category still worth

investing in if AI can do the work

instead? The clock isn't counting down

to one specific product launch, it's

counting down to when investors think

that companies will stop paying for so

many human seats altogether. By the way,

if you've ever wondered if a stock is

actually worth its market price, you are

not alone. Finding a company's fair

value is one of the hardest parts of

investing, but I found a tool that makes

it a lot easier. Investing Pro is on

investing.com

and it's become a core part of my

research workflow because of how

transparent it is. Their fair value

model doesn't just rely on one approach,

it combines discounted cash flows,

earnings-based models, and peer

comparisons. Then, it dynamically

weights them based on which methods

matter most for each kind of company.

But what I really like is that it shows

you exactly how each model contributes

to the final fair value, both on the

company page and right inside your

watchlist, making it super easy to spot

which stocks are undervalued and which

ones are actually overpriced. Then,

there's ProPicks AI, which builds on

that same foundation. It uses the same

multi-model logic plus machine learning

to find the best stocks for different

strategies, like value, growth, and even

dividends. And it explains why each

stock is chosen, so you can see the

reasoning behind every pick.

Investing.com is filled with tools that

I actually rely on. It saves me time and

it gives me real confidence in my

valuations. That's why I'm glad they're

sponsoring this video. But here's the

ironic part, they're running a huge

flash sale April 20th through 22nd, so

you can grab Investing Pro itself way

below fair value with my link below. All

right, so this is much bigger than just

Claude design. Software as a service is

one of the cleanest business models that

Wall Street has ever seen. A company

hires more people, those people need

more tools. More tools mean more seats

and more seats means more recurring

revenue. And because software is

delivered through the cloud, the gross

margins can be 70% or more compared to

around 30% for most physical products.

You don't need to build a factory, you

didn't need to ship another box, you

just added another user to the account.

That's why investors always pile into

new SaaS companies, but the whole engine

depends on a simple assumption.

Work is done by humans and every human

needs a seat. AI agents directly attack

that assumption. SaaS companies can add

AI features, they can rename products,

and they can talk about co-pilots,

assistants, and platforms, but if the

customer's economic reality changes from

100 workers needing 100 seats to 20

workers managing agents, then the

revenue model has to change and that's

brutal for software companies. Per seat

pricing is predictable, it's easy to

budget for and it's easy to renew.

Usage-based pricing makes more sense in

the world of AI, but it's also hard to

know what you're actually paying for.

Tokens, API calls, agent runtime, and

even harder to calculate the value. And

if software companies can't price their

products for the AI era, then they might

not be a part of it. So, the SaaS

apocalypse isn't really about AI killing

software platforms, it's about AI

killing software pricing, which is what

made them great investments. And every

new AI capability isn't just a product

launch, it's another excuse to cut more

seats. But there's an even bigger

problem for software stocks. Anthropic

doesn't have to beat these companies at

their own game. Traditional software

competition is easy to understand. One

tool is better for group collaboration,

another is faster for prototyping, or it

has the best API for developers, or it

has the best price. The products fight

feature by feature, and customers decide

which workflows they prefer. That's not

the game that Anthropic is playing.

Claude doesn't need to be the best

design tool. It doesn't need to be the

best security system, the best CRM, or

the best marketing automation platform.

It only needs to bypass the parts of the

workflows that makes companies pay for

so many seats. This is why Claude Design

is so dangerous, even if designers say

the outputs are garbage today. The real

threat isn't a senior designer suddenly

opening Claude and canceling Figma

tomorrow. It's the founder who used to

hire a freelancer for his first drafts,

the product manager who used to ask a

design team for a mock-up, the marketer

who used to open three different tools

to draft a landing page. Those people

were never Figma's highest end users,

but they were a core part of the funnel.

They created demand for more design

tools, for more design features, and

ultimately for more design seats. If

Claude Design absorbs the first part of

that funnel, the entire pricing model

for that software category starts to

break. The same for cybersecurity.

Mythos doesn't replace CrowdStrike as an

endpoint detection and response

platform. It just needs to show that

vulnerability discovery, exploit

chaining, and patch generation can be

done without it, or at least without so

many seats. That's what makes this

disruption so different from the

previous ones. Software disrupted other

software by making the user experience

better, faster, or cheaper. AI disrupts

software by removing the user

altogether. Some software companies will

survive this AI overhaul, and they maybe

will even thrive in it. If you own the

system of record, the proprietary data,

or the permissions, agents may make your

platform more valuable by using it way

more than people ever did. But, if you

make money from humans clicking through

your workflow all day, you're in a much

worse position. And over the next 257

days, this question will rotate through

every software category on the market.

And the worst part for software stocks

is that Anthropic is not alone. Google

is pushing the same direction through

Stitch, their AI-first design tool that

lets users create, iterate, and

collaborate on UI designs using natural

language. It also has a built-in design

agent that can reason across an entire

project, and even an agent manager that

lets users work on multiple ideas in

parallel. This isn't a feature in Figma.

This is Google trying to own the entire

workflow from prompt to user interface.

OpenAI is doing something similar with

Codex. The goal isn't an AI assistant.

The goal is an autonomous system that

can write code, reason about software,

and take on bigger chunks of real

engineering work independently. Now,

extend that across the entire economy.

Every frontier AI lab wants the same

thing, to take over those expensive

workflows before legacy software

companies can turn their own platforms

into agentic systems. Finance, HR,

procurement, project management,

marketing operations, the list goes on

and on. These are not just random

features. They're the places where human

knowledge workers spend their day and

companies spend their budgets. And if AI

labs can own the user interface, then

they can route the work to whatever

tools, whatever databases, APIs, and

cloud services they want to sit below it

while capturing more of the profits. The

work still touches all those tools, but

the user's time, skill set, and even

loyalty shifts to the agent interface

that coordinates all that work for them.

A new AI product every 2 weeks means

software companies can't have a normal

product development cycle. They don't

get six quarters to copy someone else's

feature and bundle it with their other

products. The market just assumes they

can't keep up. By the time Figma

explains why Claude Design isn't good

enough, Google Stitch is already a part

of the conversation. By the time

cybersecurity companies explain why

Mythos won't replace them, the market is

already thinking about what comes after,

and what could. The threat isn't anyone

AI model. It's every major AI lab racing

to turn entire software categories into

prompts. Some software companies will

become platforms for AI agents. Others

will become silent features behind

someone else's interface. And others

still will simply lose enough paid seats

to start losing shareholders, too. But,

investors need to be careful, because

not all software is going to get

disrupted equally. The best way to find

great long-term investments and avoid

the bad ones is understanding a

company's products, not just their

profits. The market loves simple

stories. AI killed SaaS. Adobe is dead.

Cybersecurity is doomed. But, smart

investors understand that the real story

isn't that simple. The most exposed

software companies all have four things

in common. The work is repetitive. The

pricing is per seat. The industry isn't

highly regulated, and the workflows can

be described using everyday language.

Customer support, sales, project

management, data entry, design,

marketing. These are all gigantic

software categories where users spend

time and energy translating their intent

into software clicks. AI is very good at

attacking those kinds of translation

layers. The surviving software companies

look very different. They have deep

proprietary data. They require audit

trails. They carry regulatory risk. They

connect to real-world operations. They

act as systems of record. Patient

records and healthcare data. Financial

compliance systems. Cybersecurity. These

areas are harder to disrupt because the

value extends well beyond the interface.

This is why companies like CrowdStrike,

Can Paltrow Networks could end up

winning big. Even though Mythos can find

vulnerabilities and generate patches,

Mythos also creates more attack

opportunities and makes them happen

faster, which makes cybersecurity

platforms more valuable overall.

Cybersecurity isn't simply getting

killed by AI, but it is getting

rewritten by AI. Adobe is in a similar

situation. Adobe can be attacked by

Claude Design and still become a data

aggregator, a specialized partner, or

even a system of record for creative

work. That distinction is what matters

for portfolios. The danger isn't owning

software. The danger is owning software

just because it used to grow fast while

the actual underlying economics are

changing underneath you. A company can

survive as a platform and still become a

worse stock. A product can be important

and still lose its pricing power. A

market can grow while the companies on

top of it keep losing market share. AI

agents don't need to kill software

stocks, because if the value is leaving

software seats and license fees, it has

to go somewhere else. And that's where

we want to be investing. Every time AI

takes over another workflow, its compute

costs don't disappear. They move. The

value shifts from a human clicking

through a software interface to a model

generating tokens, reading context,

searching files, calling tools, writing

code, producing images, checking

outputs, and running another pass.

Claude Design needs vision, design

reasoning, code generation, storage,

data pipelines, and collaboration

infrastructure. Mythos needs enough

compute and context to search through

huge codebases, analyze code, find

exploits, and generate patches. Google

Stitch needs Gemini for inference.

OpenAI's Codex needs cloud coding

agents. The biggest insight for

investors is that the more useful these

AI systems become, the more

infrastructure they need to power them.

Stop trying to guess which SaaS

companies will survive every new AI

product launch, and start asking who

gets paid every time one happens. Claude

is hosted on Amazon Bedrock, Google

Vertex AI, and Microsoft Foundry. That

means every product they launch creates

demand for the cloud platforms

underneath it. These companies are not

hiding from the AI wave. They're

building the infrastructure, the

security, the cloud, and the enterprise

layers that the AI wave needs to scale.

The same pattern that tells you which

SaaS stocks are in trouble also tells

you which infrastructure stocks will win

big. And the more AI agents replace

workflows, the more tokens get consumed,

the more GPUs, memory, networking,

storage, packaging, power, and cloud

orchestration the entire AI industry

needs. Claude isn't killing computing.

Claude is monetizing it with work that

used to happen inside those SaaS seats.

That's where I'm investing. For me,

Nvidia is still the most obvious and the

best way to invest in that compute.

Every software workflow will eventually

become an agentic AI workload. Every

time Claude, Gemini, Codex, or an

enterprise agent does more work, the

industry needs more accelerated

computing, more networking, more racks,

and more inference-optimized systems.

Nvidia's data center business is the

center of gravity for that entire shift.

In their latest earnings, Nvidia

reported $68.1 billion of total revenue,

with $62.3 billion of that coming from

AI data centers. That means AI is over

90% of Nvidia's revenues. If Nvidia is

the compute, then Micron is the memory.

Inference today isn't compute-limited.

It's memory-limited. Long context agents

need to hold more information. Code

analysis needs to scan more files.

Design agents need multimodal context to

see mock-up images and read design

notes. The bottleneck isn't how much the

model can think. It's how much data the

model can get from memory and how fast.

Micron's latest earnings showed that

exact need. They reported a record $23.9

billion in revenue with gross margins

coming in at 74%

up from just 37% 1 year earlier. And as

agents with huge context windows replace

normal software interfaces, memory

becomes one of the biggest physical

limits of the whole system. And then

there's the cloud infrastructure

providers, Amazon, Microsoft and Google.

These hyperscalers win when enterprises

don't build their own AI infrastructure

and most enterprises don't. What do most

of them want? They want model access,

security, compliance, data connections,

developer tools, governance, uptime, and

of course billing. That points them

towards AWS, towards Azure, and towards

Google Cloud. AWS reported $35.6 billion

of revenue last quarter, up 24%

year-over-year with $12.5 billion of

operating income. Microsoft reported 39%

Azure growth and $51.5 billion of

Microsoft Cloud revenue. Google Cloud

reported $17.7 billion

of revenue in Q4 of last year. That's up

48% year-over-year with 30.1% operating

margins. The trillion-dollar companies

that most investors think are too big to

buy are actually growing faster than

some startups. That's because they're

not just renting servers. They're

becoming AI factories. They host the

models, they sell the platforms, they

secure the enterprise data, and they

absorb the capex costs that individual

software companies can't afford. And

then, powering all of that is the

semiconductor supply chain. TSMC is the

hardware factory responsible for the

software apocalypse. Advanced packaging

is what lets GPUs and high-bandwidth

memory sit close enough together to move

data at the speeds that AI needs. Every

software is dead headline is actually

powered by the chips that are built,

packaged, tested, and shipped by TSMC.

TSMC, ASML, and Broadcom are the chip

building, the packaging, and the

networking companies tying it all

together. And I'm only investing in

software companies that have proprietary

data, compliance, or deep workflows like

CrowdStrike, Palo Alto Networks, and

Palantir, not just pretty interfaces

that customers pay for by the seat. I'm

not saying every SaaS company is

automatically uninvestable, but I am

saying if a software company charges per

seat for work that an AI agent can do, I

want to understand why companies are

keeping their seats. If they claim that

AI will expand their usage, I want to

see how that turns into more revenue.

And if they say they're the system of

record, I want to see whether agents

make that system more valuable or just

hide it behind their own chat windows.

Said another way, I don't want to own

software stocks just because they

survived the last 3 years since ChatGPT

came out. I want to own the systems that

get paid every time another software

category gets disrupted by AI. And that

might keep happening faster than most

investors expect. There are 257 days

left in 2026. And if AI agents come for

every software category like Claude just

came for design and cybersecurity before

that and legal software before that, the

question isn't whether software still

matters. The question is whether

investors want to own these software

stocks or the AI systems that are

actively disrupting them. Let me know

what you think in the comments. Is

Claude design just a flashy demo or is

it part of a bigger shift from software

to AI? And if you want to see even more

science behind the stocks, check out

this video next. Either way, thanks for

watching and until next time, this is

ticker symbol U. My name is Alex

reminding you that the best investment

you can make is in you.