Category: Business

Strategy CEO Phong Le signaled on Thursday that retail investors are becoming increasingly interested in the Bitcoin-buying firm’s flagship preferred share relative to its common stock, highlighting who is exposed to the company’s shift in fundraising efforts.

Although individuals currently hold approximately 40% of the company’s ordinary shares, Lee noted in a post on X that they presently make up around 80% of those invested in STRC. Strategy started pitching the shares alongside its $2.5 billion debut last year.

At a market cap of $5 billion, Lee suggested that STRC’s popularity among retail investors indicates that they “prefer low-volatility, high-yield digital credit.” The assessment comes as Strategy’s common stock (MSTR) price has plunged 56% over the past six months to $134.

Not long after STRC debuted in July, Strategy Executive Chairman and co-founder Michael Saylor said the product that currently pays 11.5% in dividends annually could be interesting for a “whole new class of people.” Those remarks focused on investors like retirees, yet the product has also started showing up on its Bitcoin-buying peers’ balance sheets.

Platforms common among retail investors have expanded access to STRC, which trades on the Nasdaq, including Robinhood, Kraken, and Webull. At 80% of STRC’s market cap, Lee indicated that retail investors hold $4 billion worth of the dividend-paying product.

On a notional basis, that’s still less than the value of common shares that Lee said retail investors hold. A 40% slice of Strategy’s $46.3 billion market cap is currently $18.5 billion.

The notion that Strategy’s common stock is losing preference among retail investors makes sense when viewed through a risk-adjusted lens, according to Mark Palmer, an equity research analyst at investment banking firm Benchmark-StoneX.

“The company’s common stock offers theoretically unbounded upside, but it is essentially a leveraged, non-yielding Bitcoin proxy and therefore better suited for sophisticated, risk-tolerant investors,” he told Decrypt. “STRC offers a predictable return through its high-yield, low-volatility, and significant Bitcoin overcollateralization that limits downside, and as such it maps better to how most retail investors are accustomed to thinking about income-generating assets.”

Analysts at Benchmark, who have penciled in a year-end price target of $705 for Strategy, are among the Bitcoin-buying firm’s most bullish on Wall Street. Analysts at TD Cowen, for example, pared their price target to $500 from $440 earlier this year.

The investment bank’s managing director of equity research, Lance Vitzana, recently told Decrypt that STRC’s uptick in issuance followed Strategy’s annual conference in Las Vegas last month. He noted that STRC was marketed aggressively during the two-day confab.

So far this month, Strategy has raised more than $1.5 billion via the dividend-paying product, which is engineered to trade at near its $100 par value. That represents around 33% of the product’s market cap, including its multi-billion-dollar public offering.

When the preferred share trades above that threshold, Strategy issues more shares to grow its Bitcoin stockpile. If the product lingers below, then the firm has indicated that it will hike the dividend in an effort to increase demand and lift STRC back towards its target.

Even though institutional investors are allocating to STRC, Palmer said that group is unlikely to displace demand from individuals. That’s because institutions tend to prefer the relative liquidity of Strategy’s common equity and asymmetric risk-reward profile, he said.

“In that sense, STRC is carving out a distinct investor base rather than competing directly with Strategy’s common stock,” Palmer added. “Importantly, this dynamic strengthens Strategy’s ability to raise capital for bitcoin accumulation, as STRC effectively expands the company’s addressable investor base.”

Ripple has unveiled a comprehensive AI-powered security overhaul for the XRP Ledger, deploying automated testing tools and establishing a dedicated red team that it said has already uncovered more than 10 bugs in the blockchain’s codebase.

The company outlined its new strategy on Thursday, detailing how AI tools will be integrated across the XRP Ledger development lifecycle, including adversarial code scanning on every pull request and automated stress testing.

The AI-assisted red team focuses on analyzing how features interact in real-world scenarios, particularly at boundaries where legacy code meets new functionality.

Ripple said the next XRP Ledger software release will be dedicated entirely to bug fixes and improvements without introducing new features, signaling a shift toward prioritizing security over rapid feature deployment. The company also plans to require multiple independent security audits for significant protocol changes and is expanding its bug bounty program.

“XRPL has proven its reliability over more than a decade of operation. Our responsibility now is to ensure the ledger continues to meet the demands of global payments, tokenized assets, and institutional-grade financial infrastructure,” the blog post reads. “We will evolve XRPL by systematically strengthening the foundation it is built on.”

XRP was recently trading at $1.34, down 5% on the day amid a broader crypto market dip on Thursday. Stock prices are also tumbling amid uncertainty around the Iran conflict.

At that price, XRP is at its lowest price in more than two weeks, per data from CoinGecko. XRP set a new all-time high price of $3.65 last July, but has fallen 63% since.

Australia’s central bank is moving toward building the legal and market infrastructure needed for tokenized asset markets, as regulators begin coordinating on rules that could allow the products to trade at scale within the financial system.

In a speech on Tuesday, Reserve Bank of Australia Assistant Governor Brad Jones said the question was no longer whether tokenization had a future in Australia’s financial system, but how it would be implemented, following the conclusion of the bank’s Project Acacia research program into tokenized assets and money.

The RBA said it would work with other regulators and industry on a new digital market infrastructure sandbox to test tokenized assets, tokenized money, and settlement systems in a longer-term environment designed to support commercialization, rather than short-term pilot programs.

The central bank also confirmed it is coordinating with other agencies on the legal and regulatory framework for tokenized markets, including how tokenized assets are classified, how settlement finality works, and how new platforms would be licensed and supervised.

The push on tokenized markets comes as lawmakers move to bring crypto platforms and tokenized custody services under Australia’s financial-services regime, requiring firms that hold client tokens to obtain licenses and meet asset-safeguarding rules.

Industry participants say that regulatory coordination is the key step needed to move tokenized assets from pilot programs into real markets.

“Project Acacia represents a turning point,” Paul Stonham, chief commercial officer at Australian crypto exchange BTC Markets and a member of the project’s advisory group, told Decrypt.

 “The RBA’s decision to move from exploratory pilots to a longer-term, stage-gated sandbox environment signals genuine institutional commitment to making tokenized finance work in Australia, not just studying it.”

Stonham said the most significant development was the coordination now underway between the RBA, the Australian Securities and Investments Commission, and AUSTRAC to address legal and regulatory uncertainty that has limited institutional participation.

He said regulated digital asset exchanges are likely to play a central role in tokenized markets, arguing that tokenized assets will need to trade on transparent, centrally managed order books operated by licensed platforms to attract larger players.

The RBA said tokenization could improve efficiency and reduce risk in wholesale markets, particularly if tokenized assets and money can be settled on synchronized systems, and estimated the economic benefit to Australia could reach about $24 billion (US$16.6 billion) per year.

The bank also said further work would focus on settlement infrastructure, tokenized bank deposits, stablecoins, and the potential role of a wholesale central bank digital currency.

Google is done treating quantum computing as a future problem. On Tuesday, the company published a formal timeline for transitioning its entire infrastructure to post-quantum cryptography (PQC) by 2029—calling the move urgent and saying quantum frontiers “may be closer than they appear.”

“As a pioneer in both quantum and PQC, it’s our responsibility to lead by example and share an ambitious timeline,” the blog reads. “Quantum computers will pose a significant threat to current cryptographic standards, and specifically to encryption and digital signature.”

The announcement, signed by Google VP of Security Engineering Heather Adkins and Senior Cryptography Engineer Sophie Schmieg, describes the 2029 target as a response to rapid advances in quantum hardware, error correction, and factoring resource estimates.

In plain English: The machines that could theoretically crack today’s encryption are getting real, faster than expected.

Google’s warning rests on two distinct threats. The first is already happening. So-called “harvest now, decrypt later” attacks allow bad actors to steal encrypted data today and sit on it, confident they’ll be able to unlock it once quantum computers are powerful enough. That threat is present-tense. The second is future-facing: digital signatures, the cryptographic foundation of authentication across the internet, will need to be replaced before a cryptographically relevant quantum computer—a CRQC—arrives.

To lead by example, Google announced that Android 17 will integrate post-quantum digital signature protection using ML-DSA, an algorithm recently standardized by the U.S. National Institute of Standards and Technology (NIST). The company is also pushing PQC across Google Cloud and internal communications systems.

The 2029 deadline is not arbitrary. IBM has its own roadmap targeting fault-tolerant quantum systems by the same year. As both companies race toward that threshold, 2025 marked a turning point in the field—when error correction breakthroughs, new processor architectures, and a Caltech result trapping over 6,000 atomic qubits at once shifted the conversation from “if” to “when.”

What does it mean for Bitcoin?

Bitcoin runs on elliptic curve cryptography (or ECDSA signatures), the same class of math that quantum computers—running what’s known as Shor’s algorithm—could eventually reverse-engineer. That means: Given your public key, a sufficiently powerful quantum machine could derive your private key.

Normal computers would take centuries to crack something like this. Quantum computers may take that problem and turn it into something solvable in practical time.

The exposure is larger than most people realize. According to Project Eleven, a cybersecurity and crypto-focused startup working on protecting crypto from future quantum computer attacks, over 6.8 million Bitcoin—over $470 billion worth—sits in addresses that are vulnerable to quantum attacks, including coins from Bitcoin’s earliest days. A separate estimate from Ark Invest and Unchained puts roughly 35% of the total Bitcoin supply in address types theoretically vulnerable to a future quantum attack.

Google’s researchers recently found that cracking RSA encryption may require 20 times fewer quantum resources than previously estimated—a finding that compressed the security timeline for everything that relies on similar mathematical structures, Bitcoin included. Earlier estimates put the qubit count needed to crack Bitcoin at around 20 million. Researchers at Iceberg Quantum now suggest the number could fall to roughly 100,000.

Quantum computers have achieved almost a 10x growth in power in the last five years.

So, should we all panic and sell our coins? Not really—but we should pay attention.

First of all, Google isn’t saying quantum computers will break cryptography by 2029. It’s simply saying it plans to be ready before they do.

Also, Bitcoin developers are not asleep at the wheel. BIP 360, a proposal introducing a quantum-resistant address format called Pay-to-Merkle-Root, was recently merged into Bitcoin’s formal improvement repository. It doesn’t activate anything—but it starts the clock on a serious overhaul.

Jameson Lopp, co-founder of Bitcoin custody firm Casa, believes that even if quantum computers remain years away from posing a real threat, upgrading Bitcoin’s protocol and migrating billions in user funds could take five to 10 years on its own.

“Right now, we’re several orders of magnitude away from having a cryptographically relevant quantum computer, at least as far as we know,” Loop told Decrypt earlier this year. “If innovation in quantum computing continues at a similar, fairly linear rate, it’s going to take many years—probably over a decade, maybe even several decades—before we get to that point.”

Bitcoin’s decentralized governance means no single team can flip a switch. Miners, wallet developers, exchanges, and millions of individual users would all need to move simultaneously.

Google can set a 2029 deadline because it controls its own infrastructure. Bitcoin cannot. And that asymmetry is exactly what makes Google’s announcement matter for crypto—not as a death sentence, but as a hard deadline the network didn’t set for itself and can’t afford to ignore.

Google Research published TurboQuant on Wednesday, a compression algorithm that shrinks a major inference-memory bottleneck by at least 6x while maintaining zero loss in accuracy.

The paper is slated for presentation at ICLR 2026, and the reaction online was immediate.

Cloudflare CEO Matthew Prince called it Google’s DeepSeek moment. Memory stock prices, including Micron, Western Digital, and Seagate, fell on the same day.

So is it real?

Quantization efficiency is a big achievement by itself. But “zero accuracy loss” needs context.

TurboQuant targets the KV cache—the chunk of GPU memory that stores everything a language model needs to remember during a conversation.

As context windows grow toward millions of tokens, those caches balloon into hundreds of gigabytes per session. That’s the actual bottleneck. Not compute power but raw memory.

Traditional compression methods try to shrink those caches by rounding numbers down—from 32-bit floats to 16, to 8 to 4-bit integers, for example. To better understand it, think of shrinking an image from 4K, to full HD, to 720p and so. It’s easy to tell it’s the same image overall, but there’s more detail in 4K resolution.

The catch: they have to store extra “quantization constants” alongside the compressed data to keep the model from going stupid. Those constants add 1 to 2 bits per value, partially eroding the gains.

TurboQuant claims it eliminates that overhead entirely.

It does this via two sub-algorithms. PolarQuant separates magnitude from direction in vectors, and QJL (Quantized Johnson-Lindenstrauss) takes the tiny residual error left over and reduces it to a single sign bit, positive or negative, with zero stored constants.

The result, Google says, is a mathematically unbiased estimator for the attention calculations that drive transformer models.

In benchmarks using Gemma and Mistral, TurboQuant matched full-precision performance under 4x compression, including perfect retrieval accuracy on needle-in-haystack tasks up to 104,000 tokens.

For context on why those benchmarks matter, expanding a model’s usable context without quality loss has been one of the hardest problems in LLM deployment.

Now, the fine print.

“Zero accuracy loss” applies to KV cache compression during inference—not to the model’s weights. Compressing weights is a completely different, harder problem. TurboQuant doesn’t touch those.

What it compresses is the temporary memory storing mid-session attention computations, which is more forgiving because that data can theoretically be reconstructed.

There’s also the gap between a clean benchmark and a production system serving billions of requests. TurboQuant was tested on open-source models—Gemma, Mistral, Llama—not Google’s own Gemini stack at scale.

Unlike DeepSeek’s efficiency gains, which required deep architectural decisions baked in from the start, TurboQuant requires no retraining or fine-tuning and claims negligible runtime overhead. In theory, it drops straight into existing inference pipelines.

That’s the part that spooked the memory hardware sector—because if it works in production, every major AI lab runs leaner on the same GPUs they already own.

The paper goes to ICLR 2026. Until it ships in production, the “zero loss” headline stays in the lab.