- OpenTools' Newsletter
- Posts
- 🍌FLUX.2 Undercuts Google Nano Banana
🍌FLUX.2 Undercuts Google Nano Banana
PLUS: Google Challenges Nvidia | Trump's AI Moonshot
Reading time: 5 minutes
🗞️In this edition
Black Forest labs releases FLUX.2 image generation
Google's AI chips challenge Nvidia with Meta deal
Trump's genesis mission explained
Workflow Wednesday #47: AI-Powered Planning
In other AI news –
Suno gets the green light to use WMG artist voices
David Sacks push to kill AI laws blows up
Alphabet races toward four trillion as AI booms
4 must-try AI tools
Hey there,
Black Forest Labs undercut Google's image pricing by 4×. Meta's spending billions on Google TPUs to hedge against Nvidia.And Trump signed Genesis Mission DOE supercomputers plus Nvidia, Dell, HPE, and AMD partnerships for federal research.
We're committed to keeping this the sharpest AI newsletter in your inbox. No fluff, no hype. Just the moves that'll matter when you look back six months from now.
Let's get into it.
What's happening:
Black Forest Labs released FLUX.2, a new image generation and editing system with four models designed for production creative workflows. The release includes multi-reference conditioning supporting up to ten images, 4-megapixel editing, and improved text rendering.
The company released one fully open-source component under Apache 2.0: the FLUX.2 VAE (variational autoencoder). This module compresses images into latent space and reconstructs them at high resolution. Enterprises can use it in self-hosted pipelines, gaining interoperability with BFL's commercial models while avoiding vendor lock-in.
Four other models have varying licenses: FLUX.2 [Pro] and [Flex] are proprietary hosted offerings, [Dev] is an open-weight downloadable model requiring commercial license, and [Klein] (coming soon) will be Apache 2.0.
FLUX.2 [Pro] costs $0.03 per megapixel. A 1024×1024 image costs $0.030. Google's Nano Banana Pro charges $0.134 per 1K-2K image and $0.24 per 4K image, making FLUX.2 roughly 4-8× cheaper at comparable resolutions.
Why this is important:
Open-source VAE under Apache 2.0 lets enterprises standardize on transparent latent space without vendor lock-in. That's rare in commercial image generation where proprietary systems dominate.
4-8× lower pricing than Google's Nano Banana Pro at comparable quality is a significant cost advantage for high-volume workflows. At scale, that's millions in savings.
Multi-reference conditioning supporting ten images without separate modules simplifies brand-consistent asset creation. Most models handle one reference image max.
Our personal take on it at OpenTools:
Black Forest Labs is undercutting Google on price while matching quality.
$0.03 per megapixel versus Google's $0.134-0.24 per image is a dramatic pricing gap. For enterprises generating thousands of images monthly, FLUX.2 is financially compelling even if quality was slightly worse. Quality appears competitive.
The open-source VAE strategy is smart differentiation. Enterprises get standardized latent space they can audit and customize while BFL monetizes through hosted Pro/Flex tiers and Dev licenses. That's a sustainable open-core model.
Multi-reference support for up to ten images is a practical feature for brand consistency, product visualization, and storyboarding.
Competitors typically handle single references or require complex pipelines.
This positions FLUX.2 as a cost-efficient alternative to Google and Midjourney for production workflows. Whether that's enough to overcome incumbent advantage depends on distribution and enterprise sales execution.
What's happening:
Meta's in talks to spend billions on Google's tensor processing units (TPUs), according to a report. The deal would have Meta using TPUs in data centers in 2027, with potential cloud rental starting next year.
The news pushed Alphabet toward a $4 trillion market valuation while Nvidia shares dropped 4% in premarket trading.
This follows Google's deal to supply up to 1 million chips to Anthropic, which Seaport analyst Jay Goldberg called "really powerful validation" for TPUs. "A lot of people were already thinking about it, and a lot more people are probably thinking about it now."
TPUs were developed over 10 years ago specifically for AI tasks and are application-specific integrated circuits designed for discrete purposes. Unlike Nvidia's GPUs (created for graphics rendering but adapted for AI), TPUs were built from the ground up for machine learning.
"Meta's likely use of Google's TPUs shows third-party providers of large language models are likely to leverage Google as a secondary supplier of accelerator chips," said Bloomberg Intelligence analysts Mandeep Singh and Robert Biggar.
Google said it's "experiencing accelerating demand for both our custom TPUs and Nvidia GPUs; we are committed to supporting both."
Why this is important:
Meta spending billions on TPUs validates Google's chips as a viable Nvidia alternative. Meta's one of the biggest AI infrastructure spenders globally. Their commitment signals TPUs are production-ready.
Following the Anthropic deal (1 million chips) with Meta talks creates momentum. Two major customers in short succession suggest TPU adoption is accelerating beyond Google's own infrastructure.
Nvidia dropping 4% while Alphabet approaches $4T valuation shows the market reacting to competitive threat. Investors pricing in a potential shift from Nvidia monopoly to multi-vendor market.
Our personal take on it at OpenTools:
This is a diversification strategy, not replacement.
Meta's talking 2027 data centers with potential cloud rental next year. That's hedging against Nvidia dependency, not abandoning GPUs entirely. Bloomberg Intelligence calling this "secondary supplier" is accurate framing.
Anthropic validation matters. Anthropic's a frontier AI lab with demanding workloads. If TPUs work for them, they work for most enterprise use cases. Meta following Anthropic's lead is pattern recognition.
Nvidia dropping 4% is an overreaction if this is truly secondary supplier arrangement. But the market's pricing is in threat of margin compression if Google captures meaningful share. Nvidia's pricing power depends on monopoly. Competition erodes that.
Google developing TPUs for 10+ years and iterating with DeepMind's Gemini team creates specialization advantage. Custom chips optimized for specific workloads can outperform general-purpose GPUs on those tasks.
Michael Burry scrutinizing Nvidia over "circular AI deals, hardware depreciation and revenue recognition" adds context. If Nvidia's financials face questions while Google offers a validated alternative, switching risk-reward shifts.
This is the beginning of the end of Nvidia's monopoly, not an immediate threat. But two years from now, a multi-vendor market is more likely than today.
What's happening:
Trump signed an executive order Monday creating the "Genesis Mission"—a federal AI initiative designed to accelerate scientific breakthroughs by coordinating research across government agencies.
The effort taps Department of Energy national labs' supercomputing power to run AI experiments on federal datasets.
Partnerships with Nvidia, Dell, HPE, and AMD will add more compute capacity.
Officials are targeting breakthroughs in materials engineering, health sciences, and energy. The goal: shorten discovery timelines and lower costs for Americans.
Michael Kratsios, Director of White House Office of Science and Technology Policy, called it the "largest marshaling of federal scientific resources since the Apollo program."
Energy Secretary Chris Wright said Genesis will "reverse price rises that have infuriated American citizens" by making the grid more efficient and bringing more energy online.
Why this is important:
This positions AI as the government's top scientific priority, equal to the space race and Manhattan Project.
Trump's framing AI development as a national security race against China. Every recent executive order removes regulatory friction: easier data center permitting, blocking state-level AI regulations, streamlining infrastructure builds.
Genesis coordinates what's been fragmented. Federal agencies have AI efforts, but they're siloed. Pooling DOE supercomputers with private sector chips creates compute capacity most research institutions can't afford.
The energy angle matters. AI data centers strain the grid. Promising cheaper electricity through AI efficiency buys political cover for massive energy demands.
Our personal take on it at OpenTools:
This is Apollo-level rhetoric with execution questions.
The partnerships are real: Nvidia, AMD, Dell, HPE are committed. But "reversing energy price rises" through AI? That's overselling it. AI consumes energy; it doesn't magically create it.
The China framing works politically but distorts priorities. Scientific breakthroughs don't happen on political timelines. Promising cost reductions to voters while building energy-hungry infrastructure creates a credibility gap.
Blocking state regulations while centralizing federal AI policy is the real play here. Genesis gives Trump cover to preempt California, New York, and other states from setting their own AI rules.
If it delivers materials or health breakthroughs in 12-18 months, it's a win. If it's just compute capacity sitting idle while agencies figure out coordination, it's expensive theater.
This Week in Workflow Wednesday #47: AI-Powered Planning
This week, we’re showing you how to turn messy goals into clear, data-backed strategy — without disappearing into a five-hour planning session or a 40-slide deck.
Workflow #1: Build a 90-Day Growth Plan in 10 Minutes (Perplexity)
Step 1: Drop a single query into Perplexity — “Give me the current market landscape for [your industry] with competitors, trends, risks, and opportunities.” Watch it pull live data you’d normally spend half a day hunting down.
Step 2: Paste that output back in and tell Perplexity to rank your next 90-day priorities using the ICE or RICE frame……We break down this workflow (and two more ways to use AI to plan smarter and execute faster) in this week’s Workflow Wednesday.
Warner Music Group partners with Suno to offer AI likenesses of its artists – The record label has also dropped its lawsuit against Suno, which will soon allow users to make AI-generated music using voices from participating WMG artists.
David Sacks tried to kill state AI laws and it blew up in his face – A leaked executive order draft reveals the tech billionaire making a power play to become America’s AI policy gatekeeper.
Alphabet on pace to hit $4 trillion market value as AI gains momentum –The search and ad tech giant joins an elite club of Big Tech companies racing to dominate the booming technology as AI continues to captivate.
Flair - An AI design tool that helps users quickly and affordably generate high-quality marketing assets
RoomGPT - Allows users to take a picture of their room and generate a new version of their room in different themes
Interactive Mathematics - A platform that provides math lessons and an AI-powered math problem solver to help students improve their math skills
WellyBox - A tool that helps users track and manage their receipts and invoices
We're here to help you navigate AI without the hype.
What are we missing? What do you want to see more (or less) of? Hit reply and let us know. We read every message and respond to all of them.
– The OpenTools Team
How did we like this version? |
Interested in featuring your services with us? Email us at [email protected] |