Case 36｜Two Tax Nets, One Common Crack

Lately, several seemingly unrelated events have left me feeling there’s a hidden thread connecting them.

I. The Government’s Tax

The Australian government is considering another round of tax reforms: additional taxes on superannuation accounts over $3 million, potential cuts to capital gains tax concessions, and adjustments to negative gearing.
The official narrative is about “fiscal sustainability” and “intergenerational fairness.”
But on the ground, the reality sounds different. Small business owners worry about cost pass-throughs. Middle-class families are recalculating mortgages and living expenses. Young people watch their already distant homeownership dreams being pushed even further away.
No matter how noble the stated reasons, the ones who ultimately pay are almost always the same group.

II. The AI Industry’s Compute Tax

Meanwhile, in the AI world, prices are quietly rising.
Anthropic has shifted its billing model toward usage-based pricing. OpenAI has introduced higher-tier Pro plans. Domestic large model APIs have seen multiple price increases in the past year.
The justification is the same: compute costs are too high.
But developers are responding honestly — more and more are exploring open-source alternatives, local deployment, and ways to bypass these ever-increasing API bills.

III. The Unexpected Acceleration from Chip Bans

Jensen Huang recently made a notable comment: export restrictions will not stop China’s AI development; instead, they may accelerate the rise of stronger local competitors.
When you can’t access the most advanced chips, being forced to build your own often leads people to move faster and further.
This is not speculation — it’s already happening. Chinese AI companies have begun heavily using Huawei’s Ascend chips, at costs significantly lower than NVIDIA’s solutions. The wall intended to contain them has instead taught those inside how to make their own bricks.

IV. One Common Crack

When you place these three events side by side, a clear pattern emerges:
Whenever those above try to impose taxes or restrictions, those below begin searching for alternatives and ways to accelerate self-reliance.
Government raises taxes → businesses and individuals look for ways to avoid or shift the burden.
AI companies impose compute taxes → developers turn to open source and local solutions.
Chip bans → the restricted side accelerates its own development.
This is not conspiracy theory. It is simply the natural counter-reaction of markets and human behavior under pressure.

V. Some People Are Already Moving

I’ve noticed more and more people quietly taking action instead of just complaining:
Some are shifting focus from paid advertising to genuine reputation building.
Some are gradually moving core operations from the cloud back to local hardware.
Some have started training their own lightweight models, simply to avoid being continuously drained by platforms.
These moves are still small, but they all point in the same direction — carving out a piece of ground outside a system that keeps extracting value.

Finally, I’ll leave you with an open question:
When both “policy taxes” and “compute taxes” land on your table at the same time,
will you choose to keep paying, or start looking for a shovel of your own?

Case 36｜两張税網，同一道裂缝

最近有幾件事，看似毫不相關，卻總讓我感覺它們指向同一道隱隱的裂縫。

一、政府的稅

澳洲政府正在討論新一輪稅改：退休金超過300萬的帳戶可能被額外徵稅、資本利得稅優惠可能縮減、負扣稅政策也面臨調整。
官方說法是為了「財政可持續」和「代際公平」。
但現實是：最終付錢的，幾乎永遠是同一群人——小生意主、中產家庭、以及還在努力買房的年輕一代。

二、AI平台的算力稅

另一邊，AI公司也在悄悄漲價。
Anthropic 調整計費模式，從固定訂閱轉向按量付費；OpenAI 推出更高階的 Pro 方案；國內大模型 API 價格一年內多次上調。
理由同樣冠冕堂皇：算力成本太高。
但開發者的反應很誠實——越來越多人開始尋找開源替代、本地部署，甚至徹底繞過這些不斷上漲的 API 帳單。

三、芯片禁令的意外加速

黃仁勳最近說了一句值得注意的話：出口限制不會阻止中國發展 AI，反而可能催生更強大的本土競爭。
當你拿不到最先進的芯片時，被迫自己造的結果，往往只會讓人走得更快更遠。
這不是猜測，而是正在發生的事。中國 AI 公司已經開始大量使用華為昇騰芯片，成本遠低於英偉達方案。原本想築起的牆，結果讓牆內的人學會了自己燒磚。

四、同一道裂缝

把這三件事放在一起，你會發現一個共同的現象：
每當上面的人想「收稅」或「設限」，下面的人就會開始尋找「替代」與「加速自救」。
政府加稅 → 企業和個人想辦法避稅或轉移
AI公司收算力稅 → 開發者轉向開源和本地化
芯片禁令 → 被限制的一方反而加快自研步伐
這不是陰謀論，只是市場和人性在壓力下的自然反應。

五、有些人已經在動了

我看到越來越多人不再只是抱怨，而是開始靜靜地做一些事：
有人把重心從廣告投放轉向真實口碑；
有人把核心業務從雲端慢慢移回本地；
有人開始自己訓練輕量模型，只為不被平台持續抽水。
這些動作還很小，但它們指向同一個方向——在被抽水的系統之外，能不能先挖一塊自己的地？

最後留一個問題給你：
當「政策稅」和「算力稅」同時落在你身上的時候，
你是選擇繼續繳，還是開始找那把屬於自己的鏟子？