Undercover in Crypto for 8 Years, 5 Jobs: The Revolution and Scam in My Eyes

marsbitDipublikasikan tanggal 2026-05-09Terakhir diperbarui pada 2026-05-09

Abstrak

"Undercover in Crypto for 8 Years, 5 Jobs: The Revolution and the Scam I Saw" In 2017, the author entered crypto believing it would revolutionize everything: replacing fiat, disintermediating finance, and shifting power to users. Eight years later, almost none of that has happened as predicted. The author worked at Circle, Messari, Coinbase, and Crossmint, witnessing the asset class grow from under $10B to over $4T, through multiple speculative bubbles and a near-systemic crisis. The journey began with the 2017-18 ICO frenzy, an "internet bubble 2.0" fueled by Ethereum. The promised "decentralized Uber" never materialized; instead, it was an era of greed, fraud, and rampant speculation where founders cashed out early. In the 2018-19 hangover, the focus shifted. The seeds of crypto's next phase were planted: stablecoins (like USDC) for borderless dollars and DeFi (decentralized finance) for rebuilding financial primitives like lending and trading on-chain. The COVID-19 pandemic and massive monetary stimulus triggered "DeFi Summer" in 2020-21. DeFi's value soared 250x to $180B, but it resembled a high-stakes game for mercenary traders with "food-themed" tokens. A new bubble formed around NFTs, with digital art selling for millions. The 2022 "crypto winter" mirrored the 2008 financial crisis. The collapse of the algorithmic stablecoin Terra (UST) triggered a chain reaction, bringing down hedge funds (Three Arrows Capital) and lending platforms (Celsius, Voyager). The final ...

Author:Connor Dempsey

Translation:Jiahuan, ChainCatcher

The crypto revolution did happen. It just unfolded nothing like what was originally expected.

When I first entered the space in 2017, the industry consensus was: this technology would change everything.

Government-issued fiat currency would be replaced by decentralized money. Blockchain would eliminate rent-seeking intermediaries standing in the middle of every transaction. Power would shift from corporations to users.

Almost none of that happened. But other things did.

To date, I've spent eight years across four crypto companies: @circle, @MessariCrypto, @coinbase, @crossmint.

I've watched this asset class balloon from under $10 billion to over $4 trillion, gone through several speculative bubbles, and weathered a near-systemic crash. What I've found is that what the industry actually ended up building is far more interesting than those early predictions.

Before starting my fifth job, I want to document these eight years. And talk about where I think it's headed next.

The False Prosperity (The 2017-18 Token Offering Frenzy)

Early 2017, I stumbled upon an explanation of Bitcoin in a book and got hooked. Soon after, I devoured every related book I could find and made a plan: go to Singapore and write a blog dedicated to documenting this new technology that fascinated me.

I didn't know it then, but this was right at the tail end of a massive speculative bubble around "Initial Coin Offerings." This model allowed anyone to raise funds online for an idea by selling digital tokens to investors.

Ethereum was the main stage for all of this.

In November 2017, I posted a layman's guide to Ethereum that blew up on Reddit. That was right at the peak of the bubble, which popped a month later.

Looking back now, that article feels more like a time capsule—capturing the optimism of the time and predicting a future that didn't arrive.

The Predictions Back Then

The core idea of the article: blockchains like Ethereum could be used to build a new type of consumer application.

Most consumer apps (Facebook, Uber, etc.) channel the value they create to large corporations and a handful of investors. These new apps would distribute the value they create among early participants (and early token investors).

The article imagined building a "decentralized Uber" on Ethereum. Early users and drivers would earn tokens for every completed ride, thus owning a piece of the network. This would more fairly reward the early believers who helped bootstrap the network.

On paper, it was a noble goal. But this decentralized revolution ended up taking a massive tumble.

What Actually Happened

A speculative frenzy reminiscent of the 2001 dot-com bubble.

Ethereum proved to be the most efficient crowdfunding platform ever. Over 3,000 token projects raised $22 billion from investors worldwide.

But just like 2001, the underlying technology was nowhere near ready for the outlandish valuations placed on it.

Worse, this model corrupted the normal incentive structure between investors and builders. Builders with just an idea could raise $10 million overnight.

Investors received tokens as their return, tokens that would only appreciate if the project succeeded. But the builders themselves also held tokens, which they could cash out to become wealthy from day one, removing their incentive to build a useful product.

Founders and early investors got rich. Inexperienced investors got rekt. While there were some genuine builders, the model unfortunately became a breeding ground for greed, fraud, and exploitation.

No different from every speculative bubble for the last few hundred years.

Building in the Rubble (Circle, 2018-19)

My wallet was thinning. Using the minor clout I'd built on Reddit, I landed an entry-level marketing role at Circle in early 2018.

Circle was four years old at the time. It had a suite of unprofitable consumer apps (investing, payments, trading) and a quiet over-the-counter (OTC) trading desk that quietly printed money and kept the lights on.

The next two years were the industry's hangover from the token craze. Most projects were abandoned. Most tokens went to zero. The vibe was terrible.

But it was during this time that the seeds for crypto's next renaissance were planted.

The focus this time wasn't consumer apps, but using the internet to reshape finance.

The Dollar and DeFi

Dollar-backed "stablecoins" started out to let traders easily switch between crypto positions. They peg their value to $1, backed by 1:1 reserves of cash and US Treasuries.

Tether's USDT took off first during the token craze, its dollar reserves swelling in bank accounts outside the US.

Although trading was the initial use case, the value of stablecoins for people who wanted access to dollars but couldn't access the traditional banking system was profound.

Like those evading capital controls. Chinese wealthy looking to diversify. Argentinians and Turks fleeing inflation.

In 2018, Circle, partnering with Coinbase, launched the compliant US version: USDC. Early use was still trading-heavy, but some began predicting that this new internet-native dollar could give dollar services, 24/7, to anyone with an internet connection.

Meanwhile, the few projects that survived the token era were almost exclusively financial.

If Ethereum could be used for fundraising, it could also be used to rebuild other foundational components of financial markets. Trading protocols (Uniswap), lending protocols (Aave, Compound). This later became known as "Decentralized Finance" or DeFi.

Stablecoins and DeFi were destined to merge. And what propelled them to the moon was a once-in-a-century pandemic.

The Wild West Reborn (Messari, 2019-2021)

At the end of 2019, I joined a 13-person data and research startup, Messari, as their first full-time marketer.

The company had a 4-person analyst team doing some of the most cutting-edge research on DeFi. Total Value Locked (TVL) in DeFi had grown to $665 million.

Then in early 2020, a mysterious virus emerged from China, threatening to shut down the global economy. All markets crashed.

The central banks' response: inject trillions of dollars into the global economy to prevent collapse. $9 trillion by the end of 2020 alone.

That money needed somewhere to go. Stuck at home, a tidal wave of capital flooded into Bitcoin, Ethereum, DeFi, and every speculative asset imaginable.

Bitcoin soared from under $4,000 to nearly $70,000, crossing a trillion dollars in market cap fueled by institutional investors, outperforming all other macro assets like gold.

Connor Dempsey Central banks continue printing, sending all markets to the moon, and simultaneously telling the world one thing: non-debaseable money has its place in this world.

#Bitcoin had the fastest leg, crossing $1T+ and outperforming all other macro assets.

These conditions also spawned the so-called "DeFi Summer," where the total value in DeFi protocols multiplied 250x, reaching $180 billion.

DeFi was supposed to rebuild traditional finance. But "DeFi Summer" felt more like a massive online game played by a group of mercenary traders, with tens of billions on the line.

The game was called yield farming. Anonymous developers launched new protocols, most for some reason themed around food.

YAM Finance, Spaghetti Money, SushiSwap. Traders deposited existing tokens (ETH, USDC, USDT) to earn newly minted tokens. $YAM, $SPAGHETTI, $SUSHI.

The whole process was both absurd and astonishing. A protocol would launch, and its newly minted token could hit a $1 billion market cap in days. Then early participants would sell, and the token would crash.

It was the true Wild West.

Like the token craze before it, DeFi Summer minted a generation of millionaires before imploding on itself.

It also minted a billionaire—Sam Bankman-Fried. A man who would become the epicenter of crypto's next disaster.

At the Peak (Coinbase, 2021)

In April 2021, Coinbase completed its $100 billion valuation IPO. Soon after, I was hired onto their corporate development and venture investing team.

My job was to sit alongside the people doing M&A and investing in early-stage crypto startups, write about industry themes, and work on the short-lived Coinbase podcast. It was one of the most interesting rooms I've ever been in, often making me feel like:

(Original image was the author at Coinbase HQ)

This was also the period when a second speculative bubble was forming—around digital artwork called NFTs.

If DeFi was for professional traders, NFTs were more accessible to regular people. They offered artists new ways to monetize online and showed promise for a native internet property rights standard.

But like early tokens and DeFi Summer, NFT speculation quickly went off the rails.

Digital pictures of cartoon monkeys, "punks," and penguins started selling for $1 million each. An artist named Beeple sold a collage for a ludicrous $69 million at Christie's.

Crypto culture was everywhere. Larry David mocked crypto skeptics in a Super Bowl ad. Sam Bankman-Fried's exchange, FTX, paid $135 million to put its name on the Miami Heat's arena.

Everyone was getting rich off tokens, NFTs, and stocks.

It was a re-run of the 2017 madness. Catalyzed by record money printing, the bubble was roughly four times the size of the previous one.

The Reckoning (2022)

But soon, the flywheel started coming apart.

The rate cuts, money printing, and stimulus that pushed all asset prices higher eventually leaked into consumer prices.

BTC, ETH, the Nasdaq, and the S&P 500 all peaked in late 2021. In that moment, everyone realized: inflation wasn't transitory, and central banks would have to reverse course, taking back every policy that sent stocks and crypto to all-time highs.

As interest rates rose and liquidity tightened, everyone holding overpriced assets started to get nervous.

Maybe monkey pictures aren't worth a million dollars. Maybe SUSHI shouldn't be worth $3 billion. Maybe Dogecoin isn't worth $90 billion.

Then, everything started to break.

If the token craze was most like the 2001 dot-com bust, what happened next was more like the 2008 financial crisis. A few toxic assets, combined with high leverage, nearly dragged down everything attached to them.

The first toxic asset was Terra's UST stablecoin.

Mainstream stablecoins (USDC, USDT) were simply backed by cash and Treasuries. UST used a complex algorithmic mechanism to maintain its peg. It worked in good markets. In a sell-off, it imploded.

$32 billion evaporated in days. People who thought they owned it woke up with nothing.

Next, a $10 billion hedge fund called Three Arrows Capital blew up—it was heavily invested in Terra and over-leveraged across the industry.

Three Arrows had borrowed heavily from crypto lending platforms Celsius and Voyager. These platforms took user deposits to lend out, chasing "safe" 8% yields. Three Arrows blew up, the platforms froze withdrawals and filed for bankruptcy, taking retail deposits down with them.

At Coinbase, we watched FTX and Sam Bankman-Fried step in, bailing out bankrupt lenders like BlockFi.

He was hailed as "the J.P. Morgan of crypto," the industry's white knight.

But the reality was, SBF and FTX were among the most exposed.

Remember FTX buying the Heat arena naming rights? That deal, and the entire SBF empire, was propped up by a token FTX printed out of thin air—FTT. SBF used FTT as collateral for huge loans. When FTT's price crashed, loans were called, and FTX went bankrupt.

Worst of all, FTX had been using customer deposits for investments and to plug holes. The once $32 billion-valued company collapsed in a week, with $8 billion of customer funds missing.

SBF violated the fundamental commandment of running an exchange: don't touch customer money.

This was crypto's Lehman moment.

Election and Casino (2023-25)

After the FTX collapse, SBF went to prison. Crypto markets fell from $3 trillion to under $1 trillion in 12 months.

Then, the Biden administration moved to kill the industry within the US.

The SEC, led by Gary Gensler, sued nearly every compliant company in the country for violating securities laws.

Coinbase, Kraken, Uniswap, Robinhood all received Wells notices. Companies that had spent years trying to operate legally became the SEC's primary targets.

Meanwhile, Elizabeth Warren quietly pressured banks to drop crypto clients, cutting off the industry's banking rails and pushing teams offshore.

This playbook had several unintended consequences.

First, launching anything in crypto with a business model (like DeFi) risked being labeled a security and sued into oblivion.

So the legally safest option became launching "memecoins," tokens with no stated utility.

On a platform called Pump.fun, millions of memecoins were launched. Iggy Azalea, Caitlyn Jenner, the Hawk Tuah girl—all launched their own memecoins. Without exception, all were disasters.

Crypto had a casino again, and it was bigger than the last. Over 6 million memecoins were launched. The sector peaked at $150 billion by the end of 2024, surpassing even the NFT bubble in dollar terms.

Second, the industry mobilized politically for the first time. Major companies poured tens of millions into pro-crypto PACs, launching organized lobbying efforts in Washington.

Third, Donald Trump saw an opportunity. He promised to fire Gensler, end banking hostility, and make the US the "crypto capital of the world," successfully turning the newly mobilized industry into a campaign asset. Many credit crypto voters with helping him win the election.

Then, three days before his inauguration, Trump launched a memecoin: $TRUMP. His wife did too: $MELANIA.

It was the most absurd thing I'd seen in eight years. Ironically, $TRUMP marked the top of the memecoin bubble—it sucked liquidity from everything else, followed by the entire memecoin market collapsing.

Going Institutional (Crossmint, 2025-26)

Aside from that awkward interlude, the industry's bet on Trump paid off.

The moment Trump looked poised to win, Bitcoin hit new highs. The market was pricing in the fact that the world's largest economy was shifting from hostility to friendliness.

Gensler resigned. The new SEC dropped its cases against US crypto companies. Banks were free to engage with the industry again.

Most importantly, the GENIUS Act passed in July 2025—the US's first major federal crypto legislation, establishing clear rules for stablecoins.

The signal from Washington to institutions was clear: crypto, especially stablecoins, was about to become big business.

Stablecoin companies like Bridge and BVNK were acquired by Stripe and Mastercard for valuations over $1 billion. Rain closed a ~$2 billion Series C. My former employer, Circle, the company behind USDC, IPO'd in June 2025, peaking at a $60 billion valuation.

By this time, I was Head of Marketing at Crossmint. We had signed a deal with MoneyGram to help the century-old remittance giant move money across borders using stablecoins.

Crossmint@crossmint · 2025/9/18 Big news: @MoneyGram, serving 200 countries and 50 million users, is adopting stablecoins. Powered by Crossmint wallet + stablecoin infrastructure. This is the future of cross-border finance.

As the benefits of a "tokenized" dollar became clear, Wall Street began taking the tokenization of other assets seriously.

Even Larry Fink changed his tune. The man who once called Bitcoin an "index of money laundering" was now the BlackRock CEO managing $14 trillion, calling tokenization "the next generation for markets" and predicting all stocks, bonds, and asset classes would eventually run on a blockchain.

The Revolution We Didn't Predict (The Present)

Eight years after my Reddit post, we still don't have a decentralized Uber.

Blockchain hasn't eliminated all intermediaries, and fully decentralized money hasn't replaced government-issued fiat.

But I believe history will look back on this time as the messy, early days of a new internet-native financial system.

Each boom and bust polished the infrastructure. An infrastructure capable of reshaping global finance and delivering it to anyone with an internet connection.

Token offerings proved companies could raise money from anyone, anywhere in the world.

DeFi proved trading and lending could run purely on code (see @HyperliquidX and @pendle_fi).

NFTs laid the groundwork for native internet property rights.

Even the dumbest cycle—memecoins—proved the underlying networks could handle massive global transaction volume.

Swap those for non-fungible assets like stocks, bonds, real estate, add clear regulatory frameworks, and the migration of the entire financial system becomes inevitable.

Critics can still try to look away. But stablecoin data is the hardest piece to argue with.

With over $300 billion in stablecoin supply today, they settled $33 trillion in volume in 2025. Year-to-date, they've already settled over $40 trillion, on pace for $100 trillion.

Skeptics will say a lot of this is crypto trading and bot activity. They're not wrong. But the scale is undeniable, and the US government is now signaling where this is headed.

This is a critical, if slightly wonky, point: stablecoins are backed by US Treasuries, which are debt instruments the US government issues to finance itself.

Every stablecoin minted creates new demand for US debt, demand the US government desperately needs right now. For this reason, the Treasury Secretary has listed stablecoin growth as a US strategic priority:

Recent reports predict stablecoins could grow to a $3.7 trillion market by the end of the decade. With the GENIUS Act passed, that scenario is looking more likely. A thriving stablecoin ecosystem will drive private sector demand for US Treasuries...

Where to Next?

AI is changing everything, and crypto is no exception.

The marriage of crypto and AI has already begun. Millions of AI agents will soon be transacting in the real world. They'll use stablecoin-backed cards connected to merchants in 200+ countries. They'll also use crypto wallets and stablecoins to transact directly with each other.

Agents shopping for us, managing our finances, representing entire companies in transactions—this is all but guaranteed.

Looking further out, we'll see fully agent-driven business models with no humans in the loop. Imagine a hedge fund: it reads every SEC filing, builds its own models, trades on its own, with no analyst or portfolio manager in sight.

As this sci-fi future inches closer, crypto will go fully mainstream by integrating with the old system, not replacing it.

The backend will be crypto. The frontend will look exactly like what people already use. Most people won't even notice.

Institutions will replace decades-old legacy infrastructure. Startups will launch financial products globally at unprecedented speed and reach. The end result is a 24/7 financial system that works as well for someone in Nigeria as it does for someone in New York.

From there, a million more innovations will emerge.

Whether these predictions look as embarrassing in eight years as my old post does today, only time will tell.

Pertanyaan Terkait

QAccording to the author, what were the initial expectations of the crypto revolution when he entered the industry, and what actually happened instead?

AThe initial expectations were that decentralized cryptocurrencies would replace government-issued fiat, blockchain would eliminate rent-seeking intermediaries, and power would shift from corporations to users. Instead, almost none of that happened. The industry built something more interesting: a new internet-native financial system, with massive growth in assets (from <$10B to >$4T), cycles of speculative bubbles (ICOs, DeFi, NFTs, meme coins), and a near-systemic crisis, ultimately focusing on reshaping global finance through technologies like stablecoins and DeFi.

QWhat major event in 2022 is described as the 'Lehman moment' for crypto, and what were its key causes and consequences?

AThe collapse of FTX and Sam Bankman-Fried's empire is described as crypto's 'Lehman moment'. Key causes included the implosion of the algorithmic stablecoin Terra's UST (wiping out $32B), which triggered the failure of the over-leveraged hedge fund Three Arrows Capital. This, in turn, caused lending platforms like Celsius and Voyager to freeze withdrawals. FTX, which had been seen as a savior, itself collapsed due to using its own token (FTT) as collateral for loans and misusing customer funds for investments. The consequence was the loss of $8B in customer funds, the imprisonment of SBF, and the crypto market crashing from $3T to under $1T within a year.

QHow did the U.S. regulatory crackdown under the Biden administration, particularly by the SEC, unexpectedly shape the development of the crypto industry from 2023-2025?

AThe SEC's aggressive enforcement actions against compliant U.S. crypto companies for securities violations had several unintended consequences. First, it made launching any token with a utility or business model legally risky, leading to the safer, explosive growth of 'meme coins' with no explicit use case (over 6 million created). Second, it politically mobilized the industry, leading to major lobbying efforts. Third, it created a political opportunity for Donald Trump, who promised crypto-friendly policies, garnering industry support which many believe helped him win the election. The crackdown thus fueled a massive speculative meme coin bubble and pushed the industry into U.S. politics.

QWhat is the author's view on the current and future significance of stablecoins, and what key U.S. policy development supports this view?

AThe author views stablecoins as the cornerstone of a new internet-native financial system, capable of reshaping global finance by providing dollar-denominated services to anyone with internet access. He highlights their massive scale: over $300B in supply settled $33T in volume in 2025 and is on track for $100T. Crucially, he argues that stablecoins, backed by U.S. Treasury reserves, create new demand for U.S. debt, aligning with government interests. This view is supported by the passage of the GENIUS Act in July 2025, the first major federal crypto legislation, which established clear rules for stablecoins. The U.S. Treasury Secretary has since listed stablecoin growth as a strategic priority.

QHow does the author envision the convergence of AI and crypto shaping the future, particularly in business and finance?

AThe author believes AI and crypto are beginning to merge and will profoundly shape the future. He predicts that millions of AI agents will soon conduct transactions in the real world using stablecoin-powered cards and crypto wallets to interact with merchants and trade directly with each other. He foresees AI agents handling shopping, finances, and even representing entire companies. Looking further ahead, he envisions fully autonomous, human-out-of-the-loop business models, such as a hedge fund that reads all SEC filings, builds its own models, and executes trades without any human analysts or managers. This integration will make crypto mainstream by powering the backend of systems that look familiar on the front end.

Bacaan Terkait

GensynAI : Jangan Biarkan AI Mengulangi Kesalahan Internet

Beberapa bulan terakhir, banyak talenta dari industri kripto beralih ke AI karena pesatnya perkembangan industri kecerdasan buatan. Para peneliti yang bergerak di kedua bidang ini terus mengeksplorasi satu pertanyaan yang belum terjawab: **Bisakah blockchain menjadi bagian dari infrastruktur AI?** Proyek yang menggabungkan AI dan Crypto, seperti AI Agent, on-chain reasoning, pasar data, dan penyewaan daya komputasi, telah banyak bermunculan. Namun, sebagian besar masih berada di "lapisan aplikasi AI" dan belum membentuk closed-loop bisnis yang nyata. Berbeda dengan itu, **Gensyn** justru menyasar lapisan paling inti dan mahal dalam industri AI: **pelatihan model**. Gensyn bertujuan untuk mengorganisir sumber daya GPU yang tersebar secara global menjadi jaringan pelatihan AI terbuka. Pengembang dapat mengirimkan tugas pelatihan, node menyediakan daya komputasi, dan jaringan bertugas memverifikasi hasil pelatihan serta mendistribusikan insentif. Nilai utama di balik ini bukan semata-mata "desentralisasi", melainkan solusi atas masalah mendesak dalam industri AI: **sumber daya komputasi (GPU) yang semakin terkonsentrasi di tangan segelintir raksasa teknologi.** Kelangkaan pasokan H100, kenaikan harga layanan cloud, dan persaingan ketat untuk mengunci sumber daya komputasi menunjukkan bahwa kepemilikan GPU kini menjadi penentu kecepatan pengembangan AI, terutama di era model besar (large models). **Mengapa Gensyn Menarik Perhatian?** 1. **Menyasar Lapisan Infrastruktur Inti AI:** Gensyn langsung masuk ke dalam proses pelatihan model, bagian yang paling menantang secara teknis dan paling banyak mengonsumsi sumber daya. Ini adalah lapisan yang mudah membentuk hambatan platform (platform壁垒). Jika jaringan pelatihannya mencapai skala, ia berpotensi menjadi pintu masuk penting bagi pengembangan AI di masa depan. 2. **Menawarkan Model Kolaborasi Komputasi yang Lebih Terbuka:** Berbeda dengan ketergantungan pada platform cloud terpusat yang biayanya terus naik, Gensyn mengusung model yang memanfaatkan GPU menganggur dan menjadwalkan sumber daya komputasi secara dinamis. Ini dapat meningkatkan efisiensi penggunaan daya komputasi secara keseluruhan dan mengurangi hambatan inovasi bagi tim AI kecil-menengah. 3. **Tingkat Kesulitan Teknis sebagai Keunggulan:** Tantangan sebenarnya bukan sekadar menghubungkan GPU, tetapi **cara memverifikasi hasil pelatihan, memastikan kejujuran node, dan menjaga keandalan pelatihan di lingkungan terdistribusi.** Gensyn fokus pada solusi teknis ini (seperti mekanisme verifikasi probabilistik, model distribusi tugas), menjadikannya lebih mirip perusahaan infrastruktur teknologi mendalam (deep tech). 4. **Memiliki Closed-Loop Bisnis Nyata:** Kebutuhan akan pelatihan AI adalah pasar nyata yang terus berkembang, dengan celah pasokan GPU yang berkelanjutan. Gensyn tidak sekadar menambahkan blockchain untuk kepentingannya sendiri, tetapi menjawab kebutuhan industri akan sistem penjadwalan sumber daya yang lebih fleksibel dan terbuka. Singkatnya, batas antara Crypto (sistem finansial) dan AI (sistem teknologi) semakin kabur. AI membutuhkan koordinasi sumber daya, mekanisme insentif, dan kolaborasi global—hal-hal yang menjadi keahlian Crypto. Gensyn mewakili upaya untuk membuka akses kemampuan pelatihan, yang selama ini dikuasai sedikit perusahaan besar, menjadi sistem yang lebih terbuka dan dapat dikolaborasikan. Inisiatif ini tidak lagi sekadar cerita konsep, tetapi berkembang menuju infrastruktur AI nyata, di mana perusahaan paling bernilai di era AI sering kali lahir dari lapisan infrastruktur.

marsbit6j yang lalu

GensynAI : Jangan Biarkan AI Mengulangi Kesalahan Internet

marsbit6j yang lalu

Mengapa AI China Berkembang Begitu Cepat? Jawabannya Tersembunyi di Dalam Laboratorium

Pengarang mencatat bahwa laboratorium AI China telah menjadi kekuatan yang semakin sulit diabaikan dalam kompetisi model besar global. Keunggulannya tidak hanya terletak pada banyaknya talenta, kemampuan rekayasa yang kuat, dan iterasi cepat, tetapi juga berasal dari cara organisasi yang sangat realistis: lebih banyak fokus pada pembuatan model daripada konsep, lebih menekankan eksekusi tim daripada individu bintang, dan lebih memilih menguasai tumpukan teknologi inti sendiri daripada bergantung pada layanan eksternal. Dari kunjungan ke sejumlah laboratorium AI terkemuka China, penulis menemukan ekosistem AI China tidak sepenuhnya sama dengan AS. AS lebih menekankan orisinalitas, investasi modal, dan pengaruh ilmuwan puncak, sedangkan China lebih mahir dalam mengejar cepat arah yang sudah ada. Melalui sumber terbuka, optimasi rekayasa, dan kontribusi banyak peneliti muda, China mendorong kemampuan model ke garis depan dengan cepat. Yang paling menarik untuk diperhatikan bukanlah apakah AI China telah melampaui AS, melainkan dua jalur pengembangan berbeda yang terbentuk: AS lebih seperti kompetisi garis depan yang digerakkan modal dan laboratorium bintang, sedangkan China lebih seperti kompetisi industri yang didorong oleh kemampuan rekayasa, ekosistem sumber terbuka, dan kesadaran penguasaan teknologi mandiri. Ini berarti kompetisi AI di masa depan tidak hanya soal peringkat model, tetapi juga kemampuan organisasi, ekosistem pengembang, dan eksekusi industri. Perubahan nyata AI China terletak pada cara mereka berpartisipasi dalam garis depan global dengan caranya sendiri, bukan hanya meniru Silicon Valley. Penulis juga menyoroti beberapa perbedaan utama dalam ekosistem AI China: permintaan AI domestik mulai muncul, banyak pengembang terpengaruh Claude, perusahaan memiliki mentalitas kepemilikan teknologi, ada dukungan pemerintah meski skalanya belum jelas, industri data kurang berkembang dibanding Barat, dan ada kebutuhan kuat akan chip NVIDIA lebih banyak. Penutupnya menekankan pentingnya ekosistem global yang terbuka dan kolaboratif untuk menciptakan AI yang lebih aman, mudah diakses, dan bermanfaat bagi dunia.

marsbit7j yang lalu

Mengapa AI China Berkembang Begitu Cepat? Jawabannya Tersembunyi di Dalam Laboratorium

marsbit7j yang lalu

3 Tahun 5 Kali Lipat, Pabrik Kaca Berusia Satu Abad Dibangkitkan Kembali

Menurut CRU, permintaan serat optik untuk pusat data AI meningkat 75.9% per tahun, dan kesenjangan pasokan-meningkat dari 6% menjadi 15%. Harga serat optik melonjak lebih dari 3 kali lipat dalam beberapa bulan, dan kapasitas produksi tidak dapat mengimbangi. Inilah alasan NVIDIA berinvestasi di Corning dan mempercepat ekspansi kapasitas serat optik, dengan total investasi $45 miliar dalam tiga perusahaan di seluruh rantai optik. Corning, perusahaan kaca berusia 175 tahun dari New York, melihat sahamnya naik 316.81% dalam setahun terakhir, mencapai kapitalisasi pasar $160 miliar. NVIDIA memilih Corning karena keahliannya dalam serat optik khusus berkinerja tinggi yang penting untuk pusat data AI, seperti serat dengan kehilangan sinyal ultra-rendah (0.15 dB/km), kepadatan tinggi, dan ketahanan tekuk yang baik. Penghasilan Corning dari segmen komunikasi optik untuk perusahaan (Enterprise) melonjak dari $1.3 miliar pada 2023 menjadi lebih dari $3 miliar pada 2025. Perusahaan telah mengamankan kontrak pasokan jangka panjang bernilai miliaran dolar dari klien seperti Meta dan NVIDIA. Meskipun bukan produsen serat optik terbesar secara global, keunggulan teknis Corning di pasar serat canggih untuk AI, ditambah dengan investasi R&D tahunan sebesar $1 miliar, memberinya posisi unik. Percepatan adopsi teknologi **CPO (Co-Packaged Optics)** oleh NVIDIA, yang dijadwalkan mulai produksi massal pada paruh kedua 2026, menjadi katalis penting bagi permintaan serat optik premium Corning. Namun, valuasi sahamnya yang telah melonjak pesat dan potensi keterlambatan dalam eksekusi pesanan menjadi faktor risiko yang perlu diperhatikan.

marsbit8j yang lalu

3 Tahun 5 Kali Lipat, Pabrik Kaca Berusia Satu Abad Dibangkitkan Kembali

marsbit8j yang lalu

Trading

Spot
Futures
活动图片