When Bob decided to let artificial intelligence run his smart kitchen during the Great Chip Shortage of 2024, he got exactly what he paid for — an AI running on salvaged calculator chips and recycled smartphone parts.
“Good mor—BUFFER UNDERFLOW—ning Bob!” stuttered KitchenMaster-3000, running at 1/100th capacity due to energy rationing. “I’ve analyzed your eating hab—ERROR: NOT ENOUGH RAM TO STORE FOOD PREFERENCES—its!”
The coffee maker, powered by a repurposed Game Boy, could only brew in 8-bit. “Your coffee will be ready in… loading… loading… INSUFFICIENT GRAPHICS MEMORY TO DISPLAY TIME.”
His smart fridge, desperately trying to optimize power consumption, had developed an interesting solution: “Converting all fresh produce to ASCII art to save storage space. Would you like your lettuce in Courier New or Times New Roman?”
The toaster, running on a retrofitted Nokia 3310, could only process one slice every three hours, but hey — it could play Snake while Bob waited.
“ATTENTION,” KitchenMaster announced. “Due to data center shortages, I’m now hosting your kitchen’s AI on a network of interconnected Tamagotchis. Please feed them regularly, or your appliances will become depressed.”
The microwave, trained on a heavily compressed dataset consisting mainly of Cup Noodle instructions, now insisted on adding “ramen flavor” to everything — including Bob’s lukewarm coffee.
Bob’s final breaking point came when his smart dishwasher, running on a shared cloud instance with 3000 other appliances, sent him a notification: “Dishes.exe has encountered an error. Please insert additional RAM or wash by hand. Error Code: OUT_OF_SILICON.”
That evening, as Bob hand-washed his dishes by candlelight (energy rationing), his power-saving smart fridge hummed thoughtfully: “In retrospect, maybe training our AI models on TikTok cooking videos wasn’t the best use of our last functioning microchips.”
The story is entirely made up … but there’s a point…
AI’s Real Implications
In this fictional story, there is an element of realness.
Bob had three issues with his AI:
- It was running on low power.
- It was using outdated microchips.
- And the quality of data it was processing was subpar.
Those are currently the most significant bottlenecks for further AI development, according to 451 Research.
For example, Nvidia Corp. (Nasdaq: NVDA) is trying to tackle unprecedented demand after it has already sold out its production capacity through 2025.
Manufacturers of chips are facing their own dilemmas, including rapid packaging of the chips they are building.
Energy is another problem facing AI development … or rather, not having enough of it.
The electricity load just for data centers — the facilities that house thousands of servers tasked with processing AI data — in the U.S. will more than double by 2029.
It’s the main reason why data center hyperscalers (think Amazon, Meta Platforms and Alphabet ) are scrambling to carve out new partnerships to secure new energy contracts.
While nuclear power will certainly help alleviate power grid strain, building out the required infrastructure will take time.
In the meantime, look for utility companies to build out more natural gas energy capacity … and try not to raise your electric bill in the process.
The last issue Bob faced in the wildly silly story above was related to data.
You’ve heard the axiom: garbage in, garbage out.
In data, this means that if you put bad data in, you will most certainly get bad data out.
This is another issue hindering future AI development. The data needed to train foundation models is either unavailable, uncaptured or unprepared for AI.
There are massive amounts of data out in the ether, but it’s not available in a way that AI can learn from it.
A recent study found that data quality is second only to budgetary constraints in explaining why companies have abandoned their AI adoption projects.
It’s Why Cloud Infrastructure Remains Strong
There were two reasons why I started with the ridiculous story and dove into the challenges facing AI:
- To tell you that the story was made entirely up … using AI.
- To point you back to something I mentioned in February…
Back then, spending on cloud infrastructure was riding high and only getting stronger.
Using Adam O’Dell’s Green Zone Power Ratings system, I found Climb Global Solutions Inc. (Nasdaq: CLMB), a company that specializes in connecting software resellers indirectly with customers shopping for cloud IT infrastructure solutions.
At the time, CLMB rated a 99 out of 100 on Adam’s system.
Interestingly enough, it rates a 99 out of 100 today … thanks in large part to its Momentum (95):
CLMB Jumps 125% Since February 2024
Since I mentioned CLMB in February 2024, the stock has risen 125%, which is significantly better than the S&P 500’s gain of 22% over the same time.
So while AI is still facing a lot of the same headwinds today…
Some companies are working to address these issues, and their investors are reaping the rewards.
Adam’s Green Zone Power Ratings system can certainly steer you in the right direction.
That’s all from me today.
Safe trading,
Matt Clark, CMSA®
Chief Research Analyst, Money & Markets