IS THE 'AI BUBBLE BURST' EXACTLY WHAT WE NEED?
- candyandgrim
- 2 days ago
- 8 min read
Updated: 7 hours ago
(But it's going to financially devastate a lot of people)
"That's not loyalty. That's crisis management." Someone said this about my LLM-hopping (Grammarly → ChatGPT → Grok → Claude), and it clicked:
I'm not fickle. I'm a swing voter. I only switch when:
Current option becomes untenable (broken, expensive, obsolete)
Alternative is dramatically better - not 10% better, but leaves the old guard in the dust
Small improvements don't move the needle. Revolutions do.
Right now, the AI industry isn't building revolutions - it's building expensive copies of each other.
The bubble must burst. Not because AI is a fad (it's not), but because unsustainable business models always collapse.
And when it does? Brutal. But necessary.
THE PROBLEM: TOO MANY SAMEY PRODUCTS
Twenty-four months ago, there was ONE node-based AI workflow platform: ComfyUI (January 2023). Powerful, but UX was brutal - developers only.
Mid-2024: Weavy launched and changed everything. Node-based workflows designed for creatives, not just coders. Finally, visual AI pipelines without a computer science degree or the generative iteration roll of the dice, and a true aggregated platform.
Early 2025: Flora arrived, chasing the same market. Still fine - competition's healthy.
Then October 2025 happened.
Suddenly, everyone had nodes:
Runway
Adobe Boards
Freepik Spaces
Krea Nodes
Luvian (new player rather than a copycat - one to watch)
They all do roughly the same thing. None are 10x better. Most are panic-shipping features because Figma just paid £160M+ for Weavy, and no one wants to be left behind. Canva hasn't joined the bandwagon...yet!
This isn't innovation. This is FOMO disguised as product strategy.
And it's happening across every AI category:
AI image generators: Midjourney, DALL-E, Firefly, Ideogram, Flux, Imagen, Recraft, Stable Diffusion... (pick your poison)
AI video: Runway, Pika, Kling, Luma, Veo, Sora... (all promising the same thing)
AI coding assistants: Copilot, Cursor, Codeium, Tabnine, Cody, Replit... (how many autocomplete tools do we need?)
LLMs: ChatGPT, Claude, Gemini, Grok, DeepSeek, Perplexity, Mistral, Llama... (and 47 others)
The pattern: One company proves a market exists, 20 VC-funded clones appear within months, each claiming "better" through marginal tweaks.
WHY TOO MANY OPTIONS ARE DETRIMENTAL
History repeats - always the same story:
OPERATING SYSTEMS: Apple vs Windows vs Linux
Survivors:
Apple: Won design/creative through ecosystem lock-in
Windows: Won enterprise through ubiquity and backwards compatibility
Linux: Won servers/devs through open-source flexibility
Three survivors. Dozens of dead OSs in between.
Remember BeOS? OS/2? Solaris? Palm OS? Windows Phone? BlackBerry OS? All dead or irrelevant.
E-COMMERCE: Amazon vs Alibaba
Survivors:
Amazon: Western dominance
Alibaba: Eastern dominance
Two giants. Countless bankruptcies:
3D SOFTWARE: The bloodbath
Survivors:
Maxon (Cinema 4D): Motion design king
Blender: Open-source disruptor
Autodesk (Maya, 3DS Max): Legacy enterprise stranglehold
Three main players. But remember:
Softimage? Autodesk acquired, killed 2014
LightWave? Nearly dead, sold multiple times, irrelevant
Modo? Foundry acquired it, minimal development
Truespace? Microsoft bought it, killed it
Electric Image? Dead
Carrara? Dead
The pattern: Markets consolidate to 2-4 dominant players. The rest die or get acquired for pennies.
THE AI STARTUP DELUSION
Right now, thousands of AI startups burn VC cash, each convinced they'll be the exception.
The maths doesn't work:
OpenAI (ChatGPT)
Revenue: ~£3.2B annualised (H1 2025)
Costs: ~£10.8B burned in H1 2025
Unit economics: Loses £1.80 for every £1 earned
Survival: Microsoft life support only
Stability AI (Stable Diffusion)
CEO Emad Mostaque resigned March 2024
Company in chaos, desperately seeking buyers
Burned through funding, no clear profitability path
Fate: Likely acquisition or shutdown by 2026
Midjourney
Profitable (rare!)
Only because bootstrapped and lean (no VC burn rate)
Lesson: Sustainable model > VC hype
Most others
Burning £4-40M/year on compute costs alone
Praying for acquisition or IPO before cash runs out
90% will be dead by 2027
Reality: Being "slightly better than ChatGPT" isn't a business model. It's a death sentence.
WHY THE BUBBLE BURST IS NECESSARY (AND DESIRABLE)
1. IT FORCES REAL INNOVATION
When capital's free, companies optimise for growth at any cost (see: WeWork, Uber's money-losing years).
When capital's scarce, companies must solve real problems profitably or die.
Survivors will be:
Tools that are 10x better
Platforms with sustainable unit economics
Products solving pain points, not chasing hype
Products that are commerically viable - regulations are coming. They are always late to the game, but they are coming. That and a lot of IP lawsuits, of course.
Example: Dot-com bubble burst (2000-2002):
2. IT KILLS THE MEDIOCRE MIDDLE
Right now, creatives face decision paralysis:
"Should I learn Midjourney or Firefly?"
"Is ComfyUI better than Krea Nodes?"
"Do I invest time in Runway or Pika?"
Post-burst: 2-3 clear winners per category. Learn those. Ignore the rest.
Less choice = faster mastery = better work.
3. IT REVEALS WHO'S BUILDING VS WHO'S BULLSHITTING
Every AI startup claims:
"Ethically trained"
"Commercially safe"
"10x faster than competitors"
"Enterprise-ready"
When money runs out, truth emerges:
Who has real IP vs who's wrapping someone else's model?
Who has paying customers vs vanity metrics?
Who survives without VC drip-feed vs zombie companies?
The burst is a stress test. Only the structurally sound survive.
4. IT CLEARS COGNITIVE LOAD
Right now, creatives spend more time evaluating tools than using them:
Reading comparison articles
Testing free trials
Switching platforms when last one disappointed
Learning new interfaces every three months
This is exhausting.
Post-burst: Fewer tools, clearer leaders, less churn.
More time creating, less time tool-shopping.
POST-BURST LANDSCAPE (2027 PREDICTIONS)
LLMs
Survivors:
ChatGPT (if Microsoft keeps funding)
Claude (Anthropic's ethical moat + Google backing)
Open-source consortium (Llama, Mistral, DeepSeek, or mergers?)
Dead:
Grok (Musk's vanity project)
30+ no-name LLMs with no differentiation
AI image generation
Survivors:
Midjourney (bootstrapped, profitable, beloved - but if the world adopts EU law for international campaigns,it might get blacklisted)
Adobe Firefly (enterprise lock-in)
Open-source (Stable Diffusion forks)
Dead:
Ideogram, Flux, DALL-E (unless OpenAI pivots)
20+ VC-funded clones
AI video
Survivors:
Runway (first-mover, best UX, VC-backed)
OpenAI Sora (if they ever release it)
One or two dark horse or fringe specialists (Pika? Kling? Moonvalley? Someone we haven't heard of?)
Dead:
Luma, Veo, Gen-3 clones
15+ others that never found product-market fit
Node-based AI workflows
Survivors:
ComfyUI (open-source standard, can't be killed)
Adobe's solution (I am still waiting for full beta access)
Figma Weave (formerly Weavy) (Figma's £160M bet)
Dead:
Flora (unless they find a niche, find a partner - like Canva, or get acquired)
Everyone who rushed to ship nodes in late 2025...unless they find a very specific niche - like specialising in 3D workflows, or making a user-friendly version using the Canva ethos.
3D AI tools
Survivors:
Blender (open-source, community-driven, unstoppable)
Unreal Engine (Epic's cash, real-time dominance)
Autodesk Maya/Max (legacy enterprise lock-in, but fading)
Dead:
C4D/Maxon (unless we see an AI revolution or pricing drops. As a long-time C4D user, this makes me very sad.)
Houdini (too niche for mass market)
Dozens of AI 3D startups (unless they are absorbed)
THE UNCOMFORTABLE TRUTH
If you're building your workflow on a tool that:
Isn't profitable
Doesn't have deep-pocketed backers
Isn't open-source with community momentum
Isn't the first or unique
...you're building on sand.
That shiny new AI tool you just learned? 50% chance it's dead in 18 months.
This isn't pessimism. It's pattern recognition.
BUT HERE'S THE BRUTAL BIT: IT'S GOING TO FINANCIALLY DEVASTATE PEOPLE
WHO GETS HIT HARDEST?
1. Early employees at AI startups
Took equity instead of full salary
Company dies before IPO or acquisition
Stock options = worthless paper
Real example: Stability AI employees (2022-2023), equity now nearly worthless
2. Contractors/freelancers who specialised too early
"I'm a Midjourney expert!" → Midjourney pivots or dies
"I only use Runway!" → Runway acquired, features change
Skills become obsolete overnight
3. Small studios that invested in proprietary tools
Bought annual licences for tools that disappear
Built entire pipelines around one platform
Sunk cost fallacy becomes sunk cost reality
4. VC-backed founders
Burned through £4-16M
Failed to find product-market fit
Personal guarantees, ruined credit, bankruptcy
"Failed founder" stigma in next fundraise
5. Late-stage investors
Series B/C investors who bought at inflated valuations
Down rounds, liquidation preferences, total loss
Example: Anyone who invested in Stability AI 2023-2024
THE HUMAN COST
Mental health crisis among founders/employees as companies collapse
Career gaps for workers at failed startups (CV damage)
Regional economic hits (San Francisco, London, Tel Aviv AI hubs)
Lost productivity as creatives re-learn tools after platforms die
This isn't just "creative destruction." It's actual destruction - of livelihoods, savings, careers.
SO WHAT DO WE DO?
1. BET ON CATEGORIES, NOT COMPANIES
Don't become a "Midjourney expert" - become an AI image generation expert who can switch tools in 48 hours.
Learn the underlying logic (prompting, composition, style control), not just one interface.
2. PRIORITISE OPEN-SOURCE & INTEROPERABILITY
Open-source tools can't die - even if company folds, community forks it (Stable Diffusion, ComfyUI, Blender)
Interoperable formats survive - USD, FBX, glTF, PNG, MP4. Proprietary formats die with creators.
3. DELAY ADOPTION OF UNPROVEN TOOLS
New AI tool drops with VC hype? Wait 12 months.
If it's still alive, funded, iterating - consider it. If it pivoted, got acquired, shut down - you saved yourself the learning curve.
Exception: If it's open-source, dive in. Worst case, you fork it.
4. BUILD REDUNDANCY INTO YOUR WORKFLOW
Never depend on one AI tool for mission-critical work. My clients and I all got burnt when Metaio was purchased by Apple, leaving their campaigns dead in the water and a AR creative tool gap on the marketing until the likes of Zappar/Zapworks stepped in.
Example:
Primary: Firefly and/or Figma Weave
Backup: Midjourney
Emergency: Stable Diffusion (local install, can't be shut down)
Redundancy = resilience.
5. AVOID EQUITY-ONLY ROLES AT EARLY-STAGE AI STARTUPS
If joining an AI startup:
Demand competitive cash salary - equity's lottery tickets
Vest over four years with one-year cliff (protects you if company dies early)
Ask about runway - if they have <18 months cash, huge risk
Don't sacrifice salary for equity unless:
Company is Series B+ (proven product-market fit)
Founder has successful exits before
You can afford to lose 100% of that equity
6. DIVERSIFY YOUR SKILL STACK
Don't be "the Runway guy" or "the Midjourney specialist."
Be the "AI-augmented creative who understands workflow logic across platforms" person.
Skills that survive the bubble:
Prompt engineering fundamentals (transferable across all generative AI)
Node-based workflow thinking (ComfyUI logic works in any node system)
USD pipelines (3D interoperability standard)
Python/JavaScript basics (automate workflows, build custom tools)
THE CLOSING ARGUMENT
The AI bubble will burst. It has to.
Not because AI is a fad (it's not), but because unsustainable models always collapse.
And when it does:
The weak will die
The mediocre will get acquired for pennies
The exceptional will dominate
And we, the creatives, will finally have clarity instead of chaos.
Fewer tools. Clearer winners. Less cognitive load. More time creating.
But let's not pretend it won't hurt.
People will lose jobs. Founders will lose companies. Investors will lose millions. Entire ecosystems will collapse overnight.
The burst isn't the tragedy. The bubble was.
We shouldn't have funded 47 "ChatGPT but slightly different" startups. We shouldn't have VC-backed 20 node-based workflow tools doing the same thing. We shouldn't have convinced thousands to stake careers on tools with no profitability path.
But we did. And now we pay the price.
The survivors - the tools, companies, people - will be stronger, leaner, more focused.
But getting there is going to financially devastate a lot of good people.
So here's my advice:
If you're a creative: Build tool-agnostic skills. Prioritise open-source. Create redundancy.
If you're a founder: Find profitability or find deep-pocketed backers. Burn rate kills.
If you're an employee: Cash salary > equity. 18+ months runway or walk.
If you're an investor: Only back sustainable unit economics. Hype ≠ business model.
The burst is coming. Position yourself to survive it.
What's your take? Ready for the culling? Which AI tools will survive - and which are dead companies walking? Have you already been burned by a tool that died or pivoted?
Drop your story below. Let's prepare together.
Comments